XTHE

The AMD Radeon RX 9060 XT, especially the 16 GB version, is usually faster in standard gaming performance by about 5 to 10%. It also gives better value for high refreshing 1080p and 1440p gaming. The NVIDIA RTX 5060 stands out for ray tracing AI tasks and DLSS features, making it a better choice for creators or games that use a lot of ray tracing, but it only has 8 GB of VRAM.  

Key Comparisons (February 2026) 

• Performance: The RX 9060 XT is usually faster in standard rendering, with about a 6% lead at 1080p and a larger advantage at 1440p.  

• VRAM: The RX9060XT comes in a 16GB version, which means it will last longer and perform better with higher resolutions than the RTX 5060, which is limited to 8GB.  

• Features: The RTX 5060 is better at ray tracing and supports DLSS 4.  

• Efficiency: The RX9060XT uses less power, with a 150 W rating, compared to other 50 series cards.  

• Price/Value: The RX 9060 XT usually costs less than the 5060 and performs better, making it a great choice for gamers on a budget.  

Which one should you choose? 

• Pick the RX 9060 XT (16 GB) if you want the best standard gaming performance, high frame rates, 1440p gaming, and good value for your money.  

• Go with the RTX 5060 (8 GB) if you mostly play games with heavy ray tracing, need DLSS, or do content creation like 3D modeling or AI work.  

This generation, AMD finally has a strong mainstream option with the RX 9060 XT. While it isn’t perfect and doesn’t match the speed of the pricier RX 9070 XT, it’s the first time in years that AMD offers a real alternative to NVIDIA.  

This is partly because the RTX 5060 didn’t impress, particularly due to its small frame buffer and lackluster improvements over the previous generation. We tested both GPUs side by side, and one was clearly better.  

Two specs are important for the RTX 5060 and RX 9060 XT:  

1. VRAM capacity  

2. Power draw  

Power Draw is the easier of the two. NVIDIA took a massive lead in efficiency with its ADA-loveless GPUs in the previous generation, and the latest Blackwell cards stall out on the efficiency gains. The RX9060XT and RTX 5060 are in lock step in terms of power draw, which is surprising given their performance over the past few years. It’s even more surprising considering the RX9060XT, despite its issues, is clearly faster than the RTX 5060.  

Most people care most about VRAM capacity. The RTX 5060 only comes with 8GB of VRAM. The RX 9060 XT is available with 8GB or 16GB of memory. NVIDIA’s RTX 5060 Ti also comes in 8GB and 16GB versions, but the 16GB model costs over $500. We don’t love that AMD splits VRAM options across the same chip, but it’s clear that AMD offers a cheaper way to get 16GB of VRAM than NVIDIA.  

The extra capacity is important, too, though 8GB is enough for most games at 1080p. Our testing shows it becomes a limiting factor even at 1440p, especially in games that aren’t known for stressing VRAM. The RX 9060 may be available in two capacities, but for most people, the 16GB version is the only option.  

Both the RTX 5060 and RX 9060 XT target mainstream gamers who play at 1080p, and both have a $300 MSRP, but there’s more to it than just the price tag right now. Both AMD and NVIDIA have cards available at least price, and we found both for $300 with no extra fees. That’s good news, but AMD’s lineup is more complicated.  

The RX 9060 XT comes in 8- and 16-GB versions, and while the 8-GB model is at least priced, the 16-GB one isn’t. To justify an extra $50 for the RX 9060 XT over the RTX 5060, given AMD’s better performance and larger frame buffer at $390, however, we are talking about a 30% price increase for a 10% boost in performance, plus extra VRAM for higher resolutions. The 16GB RX 9060 XT is undoubtedly the better buy over the 6GB model, but it’s lost in limbo. It’s clearly more expensive than the RTX 5060, but it’s not as expensive as even the cheapest 8GB models of the RTX 5060 Ti.  

If you have a strict $300 budget, you’ll have better luck with the RTX 5060. Although AMD has models at list price, the split between 16GB and 8GB versions means the cheapest ones at MSRP will sell out first. The RTX 5060 has been out for a few weeks, and the prices are still at MSRP. Hopefully, the RTX 9060 XT will keep the MSRP for the 8GB model, but we don’t expect the 16GB model to.  

The RX9060 XT is faster than the RTX 5060, which makes sense given its higher price at 1080p. The other RX9060 XT leads by about 6%. This includes all tested games, even those with ray tracing. AMD has made significant improvements to ray tracing with RDNA 4, but the RX 9060 XT only matches the RTX 5060 in this area, which lowers its average performance. If you ignore ray tracing, which is tough now with games like DOOM, the dark ages, AMD’s lead is even bigger.  

At 1080P, the extra frame buffer from the RX9060 XT 16GB is not a significant factor. There are a few exceptions, like Indiana Jones and The Great Circle, where 8GB can still struggle even at 1080P. Performance between the 8GB and 16GB RX9060 XT is mostly similar at $300. AMD has about a 5% to 10% advantage over the RTX 5060, depending on the game.  

You can see this at 1440p. The RX 9060 XT jumps to a 10% average lead, and in games like Horizon Zero Dawn Remastered and Forza Motorsport, the gap widens. This clearly shows how 8GB graphics cards can’t keep up with 1080p. The RTX 5060 does not hit the extreme VRAM bottlenecks seen with 8GB cards at 4K, but the averages show it gives up some performance at greater resolutions simply because it does not have enough VRAM.  

The performance advantages of the RX 9060 XT alone are not enough to justify extra spending over the RTX 5060, even if AMD wins when both GPUs are $300. The main reason for choosing the 16 GB RX 9060 XT is that NVIDIA does not offer a 16 GB RTX 5060. You would need to upgrade to the RTX 5060 Ti, with the 16 GB version costing around $500, spending an extra $50. The 16 GB RX 9060 XT over the RTX 5060 may be steep, but it is your best option right now.  

Even if you are not limited by VRAM at 1080p, the RX9060 XT is better equipped for modern games with a 16GB frame buffer. The extra capacity lets it scale up to 1440p and makes it easier to enable features like frame generation. The RTX 5060 may have an edge with DLSS 4 and multi-frame generation (MFG), but MFG uses more VRAM, which is hard to allocate when 8GB cards are already short on capacity.  

It is hard to justify the RTX 5060 over the RX 9060 XT. That’s true for both the 8GB and 16GB models. For the 8GB models, the RTX 5060 and RX 9060 XT have the same MSRP. It is hard to justify the RTX 5060 over the RX 9060 XT. That’s true for both the 8GB and 16GB models. For the 8GB models, the RTX 5060 and RX 9060 XT have the same MSRP of $300. They are both available at that price, and the RX 9060 XT is faster, pure, and simple. Sure, the RTX 5060 has access to MFG, but that’s a much less useful feature on an 8GB graphics card that already struggles to scale up to greater resolutions and graphics settings.  

For the 16GB model, it’s a bit trickier. The RX 9060 XT with 16GB is faster, and it comes with more VRAM to scale to higher resolutions and fully leverage frame generation. It’s also around 30% more expensive. If you can find the RX 9060 XT at $350, you should buy it. Once prices pass $400, though, it’s time to start looking at alternatives.

Source:  Nvidia’s RTX 5060 is embarrassing against the RX 9060 XT

Google Gems and Microsoft Copilot agents are AI tools designed to perform automation. The best choice depends on whether your organization uses Google Drive or Microsoft 365.  

• Gemini gems are assistants in the Google Gemini ecosystem. They use Gemini 1.5 Pro and are useful for creative work, research, and analyzing text, images, audio, and video. They can process between 1 and 2 million tokens at a time.  

• Microsoft Copilot Agents are built with Copilot Studio. They help automate operations, manage Excel data, and work smoothly with Microsoft 365. They use GPT-4 and are made to improve project productivity in an organized way.  

Main Points of Comparison 

• Ecosystem and integration: Gems work best with Google Workspace tools, such as Docs, Gmail, and Drive. Agents are best suited for Microsoft 365 tools such as Word, Excel, Teams, and Outlook.  

• Customization and Functionality: Copilot Agents can connect to other systems via Power Automate to perform tasks. Gems are better for creating specific personas, tones, and knowledge bases.  

• Context window and performance: Gemini can process much larger amounts of information, up to 2 million tokens. This lets it analyze bigger documents or codebases than Copilot, which supports about 128,000 to 400,000 tokens.  

• Trends: Gems are strong at creative writing, brainstorming, and making images for videos. Copilot is better for reviewing documents, handling Excel data, and summarizing meetings.  

• Pricing: Both tools cost around $20-$30 per user per month with their enterprise plans.  

How to Decide? 

• Choose Gems if you use Microsoft 365 and have to analyze large amounts of data or focus on creative and multimedia projects.  

• Go with Copilot agents if your organization uses Microsoft 365, needs to manage tasks automatically with external data, or depends on structured document and spreadsheet workflows.  

Many businesses are exploring generative AI tools such as Microsoft’s Copilot and Google’s Gemini. However, choosing the right one for your organization can be tough, as each tool offers many features.  

Before adopting a GenAI tool, it’s important to review its technical details carefully. Comparing Copilot and Gemini is a helpful first step, especially when looking at their features, pricing, performance, and how well they fit into your existing systems.  

Microsoft Copilot vs Google Gemini: Core Features 

After OpenAI launched ChatGPT in November 2022, Microsoft introduced Copilot as a separate service called Bing Chat, which became available to everyone in May 2023 because of Microsoft’s partnership with OpenAI. Copilot uses the same large language model as ChatGPT and includes Bing search for instant information.  

Google joined the AI competition with BARD in February 2023, which was renamed Gemini in 2024 and 2025. Google improved its language models, releasing Gemini 3 in November 2025.  

Both platforms have made major updates to their core models, adding new features that businesses should review:  

• Context window is the amount of information it can hold at once, serving as its memory. With the Gemini API, developers get up to 2 million tokens, and Gemini advanced users can use up to 1 million Copilots. Language model OpenAI’s GPT-5.1 supports 400,000 tokens or about 350,000 words.  

• Integration: Gemini works with Google for search, which is more widely used and reliable than Bing. Both Google and Microsoft now support the Model Context Protocol (MCP), making it simpler to connect with MCP servers. MCP is quickly becoming the standard for GenAI integrations.  

• Multi-Modality: Both Gemini and Co-Pilot can generate images. Gemini uses Imagen 3, which is now widely available, and Co-Pilot uses OpenAI’s DALL · E 3. Both also offer video creation tools. Google has Vito 3.1, and OpenAI provides Sora 2.  

• Performance: Gemini 2.5 Pro currently outperforms Copilot on needle-in-a-haystack benchmarks, which measure how well a model can find and use small, hard-to-spot details in large texts; however, these results can change quickly as models improve.  

• Plugins: Both Gemini and Copilot provide various plugin options, and both companies now support the MCP protocol, which has become the universal standard for integrating third-party tools and data with Gen AI. However, Microsoft’s regular Copilot does not yet support MCP; it is currently restricted to Copilot Studio.  

Table 1: Comparison of Microsoft Copilot and Google Gemini’s Core Features 

Feature	Microsoft Copilot	Google Gemini
LLM	OpenAI GPT-5.1	Gemini 3 Pro
Price	19.99 per user per month for Microsoft 365 Premium  30 per user per month for Microsoft 365 Pi Copilot	$19.99 per user per month for Gemini Advanced  $30 per user per month for Gemini Enterprise
Context size	400,000 tokens, which is about 350,000 words	1 to 2 million tokens, which is approximately 750,000 to 1.5 million words.
Internet Search Integration	Bengal search	Google Search
API Support	Yes, with Co-Pilot 365 only!	Yes, with Gemini Enterprise only
Customizability	Yes, Co-Pilot AI Agents are part of Co-Pilot Studio aimed at businesses.	Yes, Gems and Agent support as part of Gemini Enterprise
Integration with Collaboration Suite	Yes, integrated in Office apps.	Yes, integrated with Google Workspace with Gemini Enterprise
Text to Image	Yes.	Yes
Text to Voice	Yes	Yes
Voice to text	Yes.	Yes.
Text to video	Yes, with Sora 2	Yes, with Veo 3.1
Notebook Feature	ES Co-Pilot Notebook Part of Business Edition	Yes, notebook, LLM
GPQA Diamond	88.1% (GPT-5.1)	91.9% (Gemini 2.5 Pro)
Code Assist Service	Separate (GitHub Co-Pilot)	Google Code Assist Standard

Features and Products on the Horizon 

Google continues to improve Gemini and has recently added new agent features. Now, users can create custom agents with Google within Google’s ecosystem. These agents can also connect to third-party services using standard protocols such as MCP.  

Microsoft and Google have added AI features to their web browsers, Microsoft Edge, and Google Chrome. These browsers now provide real-time assistance with video and audio support. Google has also expanded Gemini’s features on Android, especially on Google Pixel phones, by adding new app extensions.  

Microsoft is taking a wider approach with Co-Pilot. They are building a custom Co-Pilot feature into Windows powered by local LLMs. This setup lets developers create applications and features using different LLMs on a Windows computer. Co-Pilot and Co-Pilot Studio now also support other LLMs, including those from Anthropic, making the ecosystem more open and giving customers more choice in foundational LLMs.  

Right now, Google seems to be ahead in AI because it can train and build its own LLMs like Gemini. Google uses its own custom TPU chips for LLM training and inference. In contrast, Microsoft relies on OpenAI and other vendors, such as Anthropic, for these tasks.  

Co-Pilot Vs Gemini: Which Is Best 

It’s hard to choose between Co-Pilot and Gemini because GenAI tools keep changing. LLMs are continually improving in performance and context size, and vendors often add new features to stay competitive.  

For buyers evaluating Co-Pilot and Gemini for enterprise-wide adoption, an important decision framework based on fundamental business factors is essential. The best choice should align more closely with the organization’s existing technology environment, operational requirements, and future AI vision.  

Existing Ecosystem Considerations 

For businesses that mainly use Microsoft products, Co-Pilot is usually the easiest and most natural choice. It works directly with Microsoft 365 apps like Word, Excel, PowerPoint, Outlook, and Teams, which helps reduce deployment issues, training time, and compatibility problems.  

For businesses with an investment in Microsoft Azure or for companies using it, Co-Pilot is easy to add. Microsoft is making it simpler to connect Co-Pilot with Azure services, so users can link Co-Pilot agents to Azure data sources and use integrations that follow standards like MCP.  

Google is typically the natural choice currently. Gemini 3 is the best overall model for performance in coding agents, multimodality, and voice. It can handle more information than OpenAI’s GPTs.  

Google’s tight integration with the Google Workspace platform and Google Cloud services also leverages the existing Google ecosystem investment. Google is also currently the only cloud provider with the entire value chain, including data training and inference LLMs, on its own platform.  

Employee Workflows and Use Cases 

Companies should also think about which tool best matches their specific needs and use cases.  

• Productivity and joint effort: For activities such as creating documents, managing emails, and summarizing meetings, the tool that best fits the company’s daily workflow will deliver the highest returns. Usually, that means Co-Pilot for Microsoft 365 users and Gemini for Google Apps users.  

• Software development and DevOps: Both Copilot and Gemini are good coding assistants, but it’s important to check which one works best with your main programming languages, code repositories, and development processes. Most companies use GitHub Copilot, which can leverage both OpenAI GPT and Gemini 3, so it is not limited to a single language model.  

• Line of business (LOB) application integration: identify which vendor offers the best built-in connectors for key LOB systems, such as ERP, CRM, and HR platforms, including SAP, Salesforce, ServiceNow, or Workday. Microsoft offers the most options, with many ready-made connectors for enterprise agents.  

Long-term AI Strategy 

Building an enterprise AI strategy means thinking ahead. Companies should consider their long-term goals when choosing between Co-Pilot and Gemini.  

• Agent Development and Customization: Assess the ease of building custom GenAI agents specific to internal processes, such as an internal knowledge chatbot or procurement automation agent. The Azure and Co-Pilot stack offers strong tooling for those types of applications, while Gemini’s core model capabilities provide a powerful foundation for developers.  

Multimodality and future capabilities. Consider the organization’s future needs for AI to process more than just text. Gemini 3 is currently proficient in multimodal capabilities, processing images, video, and audio, and has a very large context window. If the long-term vision includes advanced computer vision or voice-enabled applications, this ability becomes a key differentiator.  

Vendor Lock-In and Openness: Microsoft plans to support multiple LLMs in its Co-Pilot ecosystem, but Google currently supports only its own models. Google’s model is currently the best, but that could change, so companies should think about what it means to be tied to just one set of models.

Source: Microsoft Copilot vs. Google Gemini: How do they compare? 

It’s 2025, and regular flagship phones just aren’t enough anymore.  

People want something fresh and interesting to catch their attention, and both Apple and Samsung seem to think thin phones are the answer.  

With the iPhone 17 Air set to launch later this year, Samsung has just revealed the newest addition to its Galaxy S25 lineup: the ultra-thin Galaxy S25 Edge. Its 5.8 mm titanium body highlights Samsung’s focus on design, underscoring that style is the main priority here.  

So where does the Galaxy S25 Edge fit into Samsung’s bigger story, and how does it stack up against the top-of-the-line Galaxy S25 Ultra?  

Design & Display 

The Galaxy S25 Edge stands out for its design. After years of phones getting bigger, manufacturers are now seeing if people want ultra-thin devices instead.  

The Galaxy S25 Edge is the first major phone to lead the way in this area. It is just 5.8 mm thick, making it much slimmer than most other flagship phones. For comparison, the iPhone 16 is 7.8 mm thick, and the regular Galaxy S25 is 7.2 mm.  

Besides being very thin, the Galaxy S25 Edge clearly looks like a Galaxy phone. It shares the same main design features as the rest of the S25 lineup, such as a flat frame and flat front and back panels. The corners are slightly curved, just like the Galaxy S25, S25 Plus, and the new Ultra.  

It also features a titanium frame, making it only the second Samsung phone to use this material. The other is the Galaxy S25 Ultra.  

The Galaxy S25 Edge introduces Corning’s Gorilla Glass Ceramic 2.0, which uses improved manufacturing processes to make the glass even more resistant to cracks and scratches; however, it does not have the same anti-reflective coating as the Galaxy S25 Ultra.  

A big change is visible on the back of the Galaxy S25 Edge. It has a slightly raised camera eye-land, but the main difference is that it only has 2 cameras. More details on this will follow.  

How Does the Galaxy S25 Ultra Stack Up in Comparison 

In terms of size, the Galaxy S25 Edge measures 158.2 x 75.5 x 5.84 mm, while the Galaxy S25 Ultra is 162.8 x 77.6 x 8.2 mm. The Ultra is taller, wider, and thicker because it is Samsung’s top flagship and includes more hardware.  

One example is the built-in S-Pen stylus, which the Galaxy S25 Edge does not support.  

The Galaxy S25 Edge is available in Titanium Silver, Jack Black, and Icy Blue.  

A Galaxy S25 Ultra is available in titanium grey, titanium black, titanium white silver, and titanium silver blue.  

The Galaxy S25 Edge has a 6.7-inch Dynamic AMOLED 2X screen with a 1-120 Hz refresh rate. The refresh rate is now more precise, supporting smaller increments instead of just 1 Hz steps. This could make the experience more seamless, and other devices will likely follow suit soon.  

It also features a QHD+ display with 2600 nits of peak brightness, similar to the Galaxy S25 Plus.  

The Galaxy S25 Ultra has a 6.9-inch Dynamic AMOLED 2x screen with a 1-120 Hz refresh rate, almost 500 ppi sharpness, and a peak brightness of about 2,600 nits.  

As expected, there aren’t many differences in display properties other than size. The color, temperature, peak brightness, and gamma are nearly identical.  

One unique feature the Galaxy S25 Ultra has, but the Galaxy S25 Edge doesn’t, is an anti-reflective coating on the screen, which prevents unwanted reflections from marring the user experience. It is a beneficial feature we hope every Galaxy adopts from now on, but it isn’t trickling down yet.  

Both devices include ultrasonic in-display fingerprint scanners for biometrics.  

Performance and Software 

The Galaxy S25 Edge and the Galaxy S25 Ultra share the same Snapdragon 8 Elite 4 Galaxy chip set, built on a 3nm manufacturing node. This one is a slightly overclocked version of the regular Snapdragon chip. It is a proper powerhouse that shines in both synthetic benchmarks and real-life scenarios.  

However, the peculiarities of the ultra-thin design have forced Samsung to tune up the Galaxy S25 Edge’s chip for efficiency. As the cooling solution inside lacks the physical space to disperse heat and the Galaxy S25 Ultra is larger, such an optimization is critical. Everything about the S25 Edge is designed for efficiency, so a slightly lower peak performance is to be expected.  

CPU benchmarks verify this: The Galaxy S25 Edge scores slightly lower than the Ultra Geek Bench on six single and multi-core tests, but overall performance remains strong.  

CPU Performance Benchmarks 

The 3D Mark graphics test shows similar results. The Galaxy S25 Edge performs a bit below the Ultra, but it still compares well to other phones.  

GPU Performance 

The Galaxy S25 Edge has 12GB of RAM, the same as the Ultra. This amount supports fluid multitasking and fast on-device AI processing.  

For storage, the Galaxy S25 Edge offers 256 GB or 512 GB, while the Ultra is available in 256 GB, 512 GB, or 1 TB.  

Both the Galaxy S25 Ultra and S25 Edge come with Android 15.1 UI7 and 7 years of software support. UI7 is a major update that refreshes the interface, adds new features, and improves existing ones.  

Camera 

The Galaxy S25 Edge focuses on design, so its camera is less of a priority. It has a dual camera setup: a high-resolution 200MP wide-angle and a 12MP ultra-wide. It uses the same sensor as the Ultra but with different lenses.  

The Galaxy S25 Ultra offers a comprehensive camera setup:  

• A 200 MP Wide-Angle  

• A 50 MP Ultra-Wide  

• A 50 MP Periscope with 5X Zoom  

• A 10 MP 3X Telephoto  

Up front, we get a 12MP camera.  

Battery life and Charging 

The Galaxy S25 Edge may not rank at the top in battery tests, but its battery life is still surprisingly good.  

The Galaxy S25 Edge’s limiting form factor has only allowed Samsung to fit a 3,900 mAh battery. That’s just 100 mAh less than the Galaxy S25’s battery. At the same time, the Galaxy S25 Ultra is equipped with a larger 5000 mAh battery, similar to those in most previous Galaxy Ultra flagships. It has superb endurance and punches above its weight in terms of battery life.  

Thanks to its efficient chipset, the Galaxy S25 Edge offers a solid average battery life of 6 hours and 22 minutes. In our tests, it lasted:  

• 16 hours and 40 minutes for web browsing  

• 7 hours and 44 minutes for video streaming  

• almost 10 hours for 3D gaming  

We set the screen brightness to 200 nits for all tests.  

As expected, the Galaxy S25 Ultra did better in all these tests. Still, it’s impressive that the Galaxy S25 Edge performed so well with a much smaller battery.  

For charging, the Galaxy S25 Edge supports 25W wired and 15W wireless charging, which matches what smaller Galaxy flagships have offered in recent years. The Galaxy S25 Ultra charges faster with a 45W wired charging, and its wireless charging speed is the same. Both phones take just over an hour to fully charge.  

Audio and Haptics 

Both phones have great audio, but the Galaxy S25 Ultra offers deeper bass and a richer sound, likely because of its larger speakers. You’ll enjoy listening on either device, but the Ultra has a bit more power.  

Haptics are traditionally superb on Samsung phones, so all is good here. This Galaxy S25 Ultra delivers slightly more precise taps, but the difference is imperceptible.  

Summary 

The Galaxy S25 Edge stands out for its super-thin design, while its hardware specs are less of a priority.  

The phone is just as capable as other Galaxy S25 models, but it tests whether people will be interested in a phone that focuses on design over top-tier hardware. For some, that’s enough.  

Galaxy S25 Ultra is Samsung’s top flagship, offering the best hardware available. It is the right choice if you want the highest specs. Its design is also premium, so the Ultra stands out just as much as the Edge.

Source: Galaxy S25 Edge vs Galaxy S25 Ultra: Beauty versus brawn 

Microsoft 365 Co-Pilot became available to GCC High customers in December 2025, introducing Generative AI to highly regulated US government cloud environments. GCC High aims to match the features of the commercial cloud, but stricter security rules mean it starts with fewer features for added safety.  

Below is an update on feature parity and differences as of early 2026:  

Key Parity and Availability Updates 

• Availability: Available since December 2025 in Word, Excel, PowerPoint, Outlook, and Teams  

• Data Residency: Data stays within the secure GCC High boundary (Azure Government) and is not used to train foundational language models.  

• Image capabilities: Image upload features in Microsoft 365 Co-Pilot are being added to government cloud environments starting in February 2026.  

• Agents: GCC High supports declarative agents created with the Microsoft 365 Agents Toolkit.  

Major Feature Differences and Limitations (GCC High vs Commercial) 

GCC High must follow tight data sovereignty and security rules (ITAR, DFARS), so several commercial features are disabled or unavailable:  

• Web Grounding (default): Web search is turned off by default in GCC High, unlike in commercial, to help prevent data leaks.  

• Security Co-Pilot: Not available for GCC high at this time.  

• Copilot in Teams chat: not available in GCC High right now.  

• Teams meeting: Co-Pilot is not available in GCC High.  

• Excel features: Co-Pilot in Excel with Python and the clean data feature are not available in GCC High.  

• Advanced Functions: Co-Pilot Powered Translation, Speaker Note Creation, and Slide Creation are not available in GCC High at this time.  

• Edge Sidebar: The Co-Pilot Chat sidebar is not available in the Edge Browser for GCC High users.  

• 3rd-party connectors, external connections, and 3rd-party plug-ins are not enabled by default.  

Compliance and Infrastructure 

• Compliance alignment: Supports FedRAMP, HI-NIST 800-53/171, DFARS, and ITAR.  

• Personnel: Only US-based screened personnel provide support.  

• Licensing requires specific add-on licenses similar to commercial, but runs on a separate, higher-cost infrastructure.  

Summary of Status 

As of late 2025 and early 2026, Co-Pilot in GCC High focuses on using internal organizational data from sources such as SharePoint and Exchange rather than web-based information. Core apps support fundamental tasks like summarizing and drafting, but advanced real-time web features are limited.  

Current and Upcoming Capabilities in GCC High 

Microsoft 365 Co-Pilot customers in GCC High can start using a strong set of features right away, with more arriving in Wave 2.  

Starting today, these features are available for rolling out in Microsoft 365 Co-Pilot for GCC high:  

• Co-Pilot in Microsoft 365 apps: AI features are built into Word, Excel, PowerPoint, Outlook, and Teams, so users can draft, analyze, and work together securely in the tools they already use. These features meet GCC’s high compliance and data residency requirements, with all data stored in US data centers and managed by approved US staff.  

• Premium features in Co-Pilot Chat:  

• Co-Pilot Chat can now include web grounding (default is off).  

• Co-Pilot Chat can work with uploaded files for eligible Microsoft 365 plans with a Microsoft 365 Co-Pilot license.  

• You get full work data reasoning with work IQ, deeper integration with Microsoft 365 apps, and more ways to customize for secure connections between outside and inside data and for custom workflows.  

• Admin controls help keep everything secure and compliant.  

• Learn more about how Microsoft 365 Co-Pilot and Co-Pilot Chat work together.  

• Compliance-ready AI and security controls: Co-Pilot for GCC High meets stringent requirements such as FedRAMP High, DFARS, ITAR, and CMMC. Data is secured both during movement and while at rest, and Microsoft Entra ID manages access. Microsoft Purview and Enterprise Data Protection are built in for security, governance, and policy enforcement.  

• Co-Pilot control system: IT teams get tools for management, license tracking, and compliance monitoring. Admins can set up and monitor Co-Pilot use to ensure it aligns with company policies. All controls are in the Microsoft 365 admin Center.  

• Notebooks and pages in the Microsoft 365 Co-Pilot App. Use notebooks and pages to work together in the dedicated app.  

Wave two features are planned for GCC high in the first half of 2026. These will include:  

• GPT-5 offers faster, more accurate, and context-aware answers for advanced analysis and problem-solving while abiding by GCC-HIGH’s strict compliance rules.  

• Image Generation: Coroll Less Users Create Custom Visuals from Text Prompts. Write in Chat and Microsoft 365 apps with built-in controls for secure use and sharing.  

• Code interpreter lets users run Python code securely for data analysis, visualization, and automation; supports advanced analytics and automating monotonous tasks.  

• Researcher Agent offers strong research tools that bring together insights from your organization and the web (where allowed) and create well-organized reports with sources cited.  

• Analyst Agent: works like a virtual data scientist, analyzing all types of data, identifying trends, and supporting decision-making.  

• Microsoft 365 Co-Pilot connectors column lets you connect to third-party and business data sources, so you can build secure Co-Pilot-compliant workflows that go beyond Microsoft 365.  

• Co-Pilot search in the Microsoft 365 Co-Pilot app: Use Co-Pilot search to find information intelligently and in context within the app.  

All these features will be designed to meet GCC High’s strict security compliance and data residency needs. For the latest updates, check the official Microsoft 365 Roadmap and Microsoft 365 Co-Pilot service description.  

Following Steps 

General agencies in GCC High that, through the GSA, can get big discounts under the OneGov agreement announced in September. This includes a chance to use Microsoft 365 Co-Pilot at no extra cost for up to 12 months with Microsoft 365. See the resources below for more details. Contact your account team or partner to learn more. Request a demo and get started with Microsoft 365 Co-Pilot in GCC High.

Source: Microsoft 365 Copilot is now available in GCC-High 

We are excited to introduce GPT-5.2, our most advanced modern series for professional knowledge work.  

On average, ChatGPT Enterprise users report saving 40-60 minutes each day, while heavy users save over 10 hours per week. We built GPT-5.2 to help people get even more value from their work. It’s now better at:  

• making spreadsheets  

• building presentations  

• writing code  

• analyzing images  

• understanding longer documents  

• using tools  

• coordinating complex projects  

GPT-5.2 sets new standards on many benchmarks, including GDPval, where it performs better than industry professionals on well-defined tasks across 44 different jobs.  

Companies like Notion, Box, Shopify, Harvey, and Zoom have found that GPT-5.2 excels at long-term reasoning and tool use. Databricks, Hex, and Triple Veil found it especially strong in data science and document analysis. Charley Labs, JetBrains, and Augment Code report that GPT-5.2 offers top performance in coding, with clear improvements in interactive coding, code reviews, and bug finding.  

In ChatGPT, GPT-5.2 Instant Thinking and Pro will start rolling out today for paid plans. For developers, these models are already available in the API.  

Overall, GPT-5.2 offers major improvements in general intelligence, understanding long documents, and working with images. This makes it better at handling complex real-world tasks from start to finish than any earlier model.  

Model Effectiveness 

Economically Valuable Tasks 

GPT-5.2 thinking is our best model so far for actual professional tasks on GDPval, which measures performance on specific knowledge work across 44 jobs. GPT-5.2 thinking achieved a new top score and is our first model to match or exceed human experts. It outperformed or matched top professionals in 70.9% of cases, according to expert judges. These tasks include creating presentations, spreadsheets, and similar work. GPT-5.2 completed these tasks over 11 times faster and at less than 1% of the cost of expert professionals. These estimates are based on past data, and actual speed in ChatGPT may vary.  

A GDP well judge praised one remarkable output, saying, “It is an exciting and noticeable leap in output quality. It appears to have been done by a professional company with staff. It has a surprisingly well-designed layout and advice for both deliverables, though with one, we still have some minor array errors to correct.”  

In our internal tests for Junior Investment Banking Analyst tasks like building a 3-statement model for a Fortune 500 company or developing a leveraged buyout model, GPT-5.2 thinking scored 9.3% higher per task than GPT-5.1. The average score went up from 59.1 to 68.4.  

Spreadsheets and slides made by GPT-5.2 thinking are more advanced and better formatted than before.  

To use the new Spreadsheet and Presentation features in ChatGPT, you need a Plus, Pro, Business, or Enterprise plan and must select either GPT-5.2 Thinking or Pro. Some complex results may take a few minutes to generate.  

Coding 

GPT-5.2 rethinking reached a new high score of 55.6 on SWE Bench Pro, a tough test for Real Software Engineering. While the SWE Bench verifies only Python, SWE Bench Pro supports four languages and is designed to be harder, more resistant to contamination, and more relevant to industry needs.  

On the SWE bench verified (not shown in the chart), GPT-5.2 thinking achieved a new high score of 80%.  

In daily work, this means the model can more reliably:  

• debug production code  

• handle future requests  

• Refactor large code bases  

• deliver fixes from start to finish with less manual work  

GPT-5.2 thinking also outperforms GPT-5.1 thinking in front-end software engineering. Early testers found it much better at front-end development and at operating complex or unusual UI tasks, especially those with 3D elements. This makes it a strong daily partner for engineers working across the stack.  

GPT-5.2 instant update (February 10, 2026) 

We are updating GPT-5.2 instant in ChatGPT and the API to improve reply style and quality.  

Users should notice that responses are more balanced and better fit the conversation. The model also gives clearer, more relevant answers to advice-seeking and how-to questions and puts the most important information first more often.  

Update on Thinking Time Settings for GPT-5.2 Thinking in ChatGPT (February 4, 2026) 

Jan 10, 2026: We lowered the standard and light thinking time after seeing that users prefer faster responses. During this update, the extended thinking setting for GPT-5.2 was unintentionally set to a lower value, but we have now fixed it.  

February 3, 2026: We made another small reduction to standard thinking time based on testing.  

Feb 4, 2026: We are restoring the extended thinking level for GPT-5.2 thinking to its previous setting, fixing the accidental reduction from January. Extended time is now back to its earlier level.  

We regularly adjust the default thinking time of our reasoning models. These changes result from ongoing tests to find the best balance between answer quality and user response speed.  

The thinking level toggle introduced in September 2025 gives users more options beyond standard. It lets them pick the right-thinking level for their question, whether they want quicker, lighter responses or more detailed reasoning when depth and accuracy are important.  

Thinking time is not directly comparable among different models. Each model is tuned separately to work best for users. We will keep changing these settings as models change and will continue to give users clear controls when there are important trade-offs to consider.

Source: GPT-5.2 Instant Update (February 10, 2026) 

Apple’s annual shareholder meeting is set for February 24 at 11 a.m. Eastern time. At that time, participants can vote and ask questions.  

At annual shareholder meetings, participants vote on issues raised by shareholders and review progress toward future goals. These meetings also cover the company’s performance over the past year.  

At Apple’s February 24 meeting, participants will vote to elect the board of directors and approve the independent public accounting firm for 2026. They will also vote on the non-employee director’s stock plan, executive pay, and a shareholder proposal about Apple’s manufacturing in China.  

Apple announced the 2026 shareholder meeting in an SEC filing on January 8. In the same filing, CEO Tim Cook shared a letter to shareholders spotlighting the company’s 2025 achievements, including the launches of AirPods Pro 3, the iPhone 17 lineup, and the upgraded M5 Apple Vision Pro.  

Our latest iPhone lineup features the most powerful iPhone we’ve ever made, with the iPhone 17 Pro, and the thinnest iPhone we’ve made ever made, with iPhone Air and said Cook, as always, we’ve continued to build our values into everything we do, from our devotion to protecting people’s privacy and security to designing our products to be accessible to all.  

Apple’s announcement also details Tim Cook’s 2025 target compensation, which is about $74 million. Most of this amount comes from stock options, while the annual cash payout is much smaller.  

A rumor from July 2025 suggests that Tim Cook may eventually become chairman of Apple’s board of directors after he retires. Hardware chief Jon Ternus is seen as a possible successor for the CEO role.  

The SEC filing from Thursday mainly acts as a proxy statement listing Apple’s achievements in 2025. For those who follow Apple’s finances, this information is not surprising. Shareholder meetings also tend to attract little public interest.  

However, one shareholder proposal stands out for its concerns about Apple’s international operations.  

Conservative think tank questions Apple’s manufacturing in China 

Participants at Apple’s 2026 shareholder meeting will vote on a proposal from a conservative think tank. The National Centre for Public Policy Research plans to submit a proposal about Apple’s activities in China.  

Shareholders affiliated with the organization will ask Apple’s board of directors to conduct an evaluation assessing risks and costs associated with the company’s continued entanglement with the People’s Republic of China.  

China-related proposals cite concerns about potential tariff increases and claims that China’s control over rare-earth materials could let the country weaponize shortages. It is also argued that Apple was the target of Chinese espionage actions in 2018 and 2019 related to the Apple Car Project.  

Apple’s board of directors recommends that shareholders vote against this proposal, stating that the requested report is unnecessary given that we already provide extensive information about our international operations. The proposal was also deemed highly prescriptive, as it attempts to improperly restrict Apple’s capacity to manage its ordinary business operations and strategies.  

While questions remain about Apple’s operations in China, the company’s 2025 revenue results clearly show its overall success.  

Apple’s SEC filing reports a record revenue of US$416.2 billion for the fiscal year 2025. Services grew by 14% year over year. The company also saw strong results in emerging markets like India, Latin America, and the Middle East.

Source:  Apple’s 2026 annual shareholder meeting is on February 24

On Friday, February 13th, 2026, at the Munich Security Conference, 16 companies from Africa, Asia, Europe, and North America announced the launch of the Trusted Tech Alliance (TTA). This group brings together global technology providers who share common principles for building a trusted technology stack, covering everything from connectivity to cloud infrastructure to semiconductors, software, and AI.  

This initiative marks the first time a group of global companies has worked together to address data storage issues. As the United States becomes more isolationist under President Trump, Europe and Asia are paying more attention to digital sovereignty.  

Governments are looking at new regulations and investing more at home to rely less on the United States and other foreign technology providers.  

This is the period when many governments and countries are under pressure to create stronger technology borders to focus more on their own digital sovereignty. Microsoft President Brad Smith told Reuters in an interview.  

Our companies are working together to set this high standard to make it really clear what the definition of trust is, he said.  

These principles guarantee that no matter where a supplier is from, alliance companies follow common commitments to transparency, security, and data protection. This creates trust and helps people everywhere benefit from technology.  

As technology advances faster than ever and the environment grows more complex, countries and customers want more reliable, dependable technology providers. There is also ongoing skepticism about digital technologies and their possible negative effects. In this situation, companies across the tech industry need to work jointly to address these concerns.  

By creating clear standards for Trusted Technology and Operating Principles, TTA members commit to working with governments and customers. Their goal is to build general trust in new technologies while supporting job creation and economic growth.  

The alliance unites leading companies that follow clear, verifiable practices and principles. These show that technology can be secure, reliable, and responsibly managed, no matter where it is developed or used.  

As of today, the signatories of the Trusted Alliance are:  

• Anthropic  

• ASML  

• AWS  

• Cassava Technologies  

• Cohere  

• Ericsson  

• Google Cloud  

• Hanwha Geo Platforms  

• Microsoft  

• Nokia  

• Nscale  

• NTT  

• Rapidus  

• Saab  

• SAP  

Participating companies have agreed on five key principles that define what it means to develop, deploy, operate, and work together as trusted global technology providers:  

1. Transparent corporate governance and ethical principles.  

2. Operational Openness, Secure Development, and Independent Assessment  

3. Strong supply chain and security oversight  

4. Open Cooperative Inclusive & Strong Digital Ecosystem  

5. Respect for the rule of law and data protection.  

These commitments mean companies must have strong governance and act ethically, build technology securely, and manage it responsibly from start to finish. They will also require suppliers to meet high global security standards and support an open cooperative digital environment that promotes innovation.  

The TTA plans to expand its community to include global providers dedicated to building a trusted, interoperable, and open technology stack. Together, they aim to support national and international efforts to strengthen sovereignty, resilience, and competitiveness.  

In the age of rapid technological change, collaboration among like-minded industry peers is essential to build buyer trust and realize the full economic and societal benefits of technology. We are joining the Trusted Alliance, the Trusted Tech Alliance, to reinforce our continued pledge to provide customers with trusted, secure, and dependable technology, said David Zapolsky, Chief Global Affairs & Legal Officer, Amazon.  

In the current geopolitical environment, like-minded companies must work together to protect security and advance high global standards to preserve trust in technology across borders, said Brad Smith, vice chair and president of Microsoft, based not on the nationality of the provider but on common commitments to customers. This alliance brings together leading companies around clear, verifiable principles that show technology can be secure, reliable, and responsibly operated wherever it is deployed.  

No single company or country can build a secure and trusted digital stack alone. Rather, trust and security can only be achieved together. That’s why, together with like-minded industry peers, we have launched the Trusted Tech Alliance, an initiative committed to verifiable trust practices across the digital stack, said Borje Ekholm, President and CEO of Ericsson.

Sources: 

Microsoft, Ericsson lead global tech alliance for digital trust

Global technology leaders launch Trusted Tech Alliance

Critical Details 

• Samsung has started mass production of HBM4, which offers a steady transfer speed of 11.7 Gbps and can reach up to 13 Gbps.  

• The advanced DRAM uses a 4NM logic-based DAI to boost performance, reliability, and energy efficiency for future data centers.  

• Samsung’s secure manufacturing process and strong supply chain support its HBM plans beyond HBM4.  

Samsung Electronics has announced it has started mass production of its HBM4 memory and has already shipped products to its customers. This makes Samsung the first company in the industry to reach this milestone and take an early lead in the HBM4 market.  

By using its cutting-edge sixth-generation 10nm-class DRAM process, Samsung achieved stable yields and top performance from the start of mass production, without requiring additional redesigns.  

Rather than using existing designs, Samsung chose to use the most advanced nodes, such as 1C DRAM and the 4nm logic process, for HBM4, said Sang Joon Hwang, Executive Vice President and Head of Memory Development at Samsung Electronics. By concentrating on process and design improvements, we can offer higher performance to meet our customers’ growing needs.  

Raising the Standard Toward Performance and Efficiency 

Samsung’s HBM4 runs at a steady 11.7 Gbps, about 46% faster than the industry standard of 8 Gbps and sets a new benchmark for HBM4. This is a 1.22 times increase over the previous HBM3E’s top speed of 9.6 Gbps. HBM4 can also reach up to 13 Gbps, helping to reduce data bottlenecks as AI models grow larger.  

The total memory bandwidth per stack is now 2.7 times higher than HBM3e, reaching up to 3.3 terabytes per second (TB/s).  

With 12-layer stacking, Samsung’s HBM4 comes in sizes from 24GB to 36GB. In the future, 16-layer stacking will allow for options up to 48GB, matching customer needs as they arise.  

To handle higher power use and heat from doubling data I/O from 1024 to 2048 pins, Samsung has added advanced low-power designs to the core die. HBM4 is 40% more power efficient thanks to low-voltage TSV technology and PDN optimization. It also improves thermal resistance by 10% and heat dissipation by 30% compared to HBM3e.  

With its excellence, performance, energy efficiency, and reliability, Samsung’s HBM4 helps customers get the most from their GPUs and better control their total cost of ownership.  

Comprehensive Yet Agile Production Facilities 

Samsung is moving its HBM roadmap forward by using its broad manufacturing resources, which include one of the largest DRAM production capacities and dedicated industry infrastructure. This solution helps keep the supply chain strong and ready for the unexpected rise in HBM4 demand.  

By closely coordinating its foundry and memory businesses through design technology and co-optimization (DTCO), Samsung maintains high quality and strong yields. Its in-house skills in advanced packaging also help speed up production and shorten lead times.  

Samsung plans to expand its technical partnerships, working closely with global GPU makers and hyperscalers, to develop next-generation ASICs.  

Samsung expects its HBM sales to more than triple in 2026 compared to 2025 and is already increasing its HBM4 production capacity. After HBM4 launches, HBM4E samples should be available in the second half of 2021, and custom HBM samples will be delivered to customers in 2027 as needed.

Source: Samsung Ships Industry-First Commercial HBM4 With Ultimate Performance for AI Computing 

Google has shared the dates of this year’s I/O event.  

• The event will happen on May 19 and 20.  

• We are not sure, yet what will be announced, but Android, Gemini, and other updates are likely to be featured.  

It was always expected that Google I/O would happen in 2026, and now Google has officially confirmed the dates.  

Google I/O 2026 will take place on May 19 and 20 at the Shoreline Amphitheater in Mountain View, California. The event starts at 10 a.m. PT, 1 a.m. ET, and 6 p.m. BST on May 19, or 3 a.m. AEST on May 20 in Australia.  

TechRadar will be there to cover all the announcements and share our reactions as they happen. While Google hasn’t revealed what’s coming, we’ve listed our best guesses below.  

Gemini Upgrades 

AI has been a major focus for tech companies recently, and since Google is a leader in this area, we expect upgrades to its AI tools, especially Gemini.  

Google’s save-the-date for I/O 2026 mentioned both AI and Gemini, so it’s likely they’ll have a big presence at the event.  

We can only guess what may be announced, but Gemini will become increasingly integrated into phones, making it better at handling tasks on your device. We might also see improvements to current AI features, like a smarter Gemini or more realistic AI photo edits.  

Android 17 Features 

The first beta of Android 17 is out, but there’s still more to come. Future betas and the final release may bring additional changes.  

So it’s likely we’ll hear about Android 17 features at Google I/O 2026. Google’s Save the Date event includes updates from Android, making it almost certain.  

The final release likely won’t arrive at I/O since it is expected in June, but the final beta might launch during the event. We don’t know what will be included, but Google is working on motion cues to help with motion sickness when using your phone in a moving vehicle, so that could be announced.  

Android XR is Google’s new augmented reality operating system for smart glasses and headsets. Right now, there aren’t any glasses that use it, and only the Samsung Galaxy XR headset runs it.  

We might see updates and new features for Android XR at Google I/O 2026, or announcements about new products that will use it.  

Google will likely want to generate excitement for this new product category. Even if there isn’t much new to show, Android XR will be featured.  

While Android XR is still new, Google is also developing a new operating system called Aluminum OS, which isn’t available yet. Aluminum OS is a desktop operating system that could combine Chrome OS and Android into one platform. Google has said it plans to do this, but hasn’t shared many details. The name, which uses the non-US spelling of aluminium, has appeared only in a job listing so far and might be a code name.  

The job listing said Aluminum OS will have AI at its core and will work on laptops, desktops, and tablets.  

Now that this information is public, it is the right time for Google to share more about Aluminum OS. I/O 2026 is the perfect place for an announcement, so we’ll likely hear more then.  

Aluminum OS 

Android XR has only just launched, but Google is already working on another operating system called Aluminum OS, which isn’t available yet.  

Aluminum OS is meant to be a desktop system, but it combines Chrome OS and Android into one platform. Google has said it plans to do this but hasn’t shared many details yet. So far, the name, spelled the British way, has only shown up in a job listing and could be a codename.  

The job listing said Aluminum OS will be built around AI and will work on laptops, desktops, and tablets.  

Now that this information is public, Google should share more about Aluminum OS. I/O 2026 would be the perfect place to hear more about it.  

Hardware Teasers 

Google I/O is primarily a developer conference focused on software, but we sometimes see hardware announcements as well.  

What could we see this year? There are no guarantees, but we might get another look at Xreal’s Project Aura smart glasses. These run Android XR and were shown at last year’s event, but they or a similar product still aren’t available. We could also see other Android XR devices.  

We won’t see any phones from the Google Pixel 11 line or the Pixel Watch 5 since these usually launch around August.

Source: Google I/O 2026 is official: here are 5 things to expect

The Apple M5 chip, launched in October 2025, delivers a major boost in AI performance compared to the M4. It has a new 10-core GPU, and each core includes a Neural Accelerator delivering four times the peak AI computing power.  

The M5 handles AI tasks like Stable Diffusion and LLMs up to 3.5x faster than the M4. It also offers 30% more memory bandwidth at 153GB and 30-45% faster ray-traced graphics, showing a move toward GPU-powered on-device AI.  

GPU & AI architecture: M5 vs M4 

• Neural Accelerators: The M5 has a neural accelerator in every core. Unlike the M4, this change greatly boosts performance for AI inference, image generation, and upscaling.  

• AI Performance: The M5 delivers up to 3.5x faster AI performance than the M4. For example, it is 1.8x faster at video enhancement in Topaz and 1.7x faster at Blender ray-traced rendering.  

• Memory Bandwidth: The M5 has a unified memory bandwidth of 153 GB/s, about 30% more than the M4. This increase is important for running large language models and demanding AI tasks.  

• Ray tracing and graphics: The M5 features third-generation hardware-accelerated ray tracing, delivering almost 50% faster frame rates in games and 3D rendering (such as in 3DMark Solar Bay) than the M4.  

Performance & Efficiency 

• General Compute: The M5 scores 15% higher in multi-threaded tasks and 10% higher in single-threaded tasks than the M4.  

• Efficiency Cologne – even with its performance improvements, the M5 keeps battery life similar to the M4, using 17-20 watts of power. However, it can get hotter than the M4 during long periods of heavy use.  

Key difference: While the M4 is still a strong chip, the M5 is designed for people who depend on AI-heavy tasks like running local language models, advanced image generation, and 3D rendering.  

Has announced the M5 chip, which brings major improvements in AI performance and upgrades across the board. Built with third-generation 3nm technology, the M5 features a new 10-core GPU with a neural accelerator in every core. This setup enables AI workloads on the GPU to run much faster, offering over 4x the peak GPU compute performance compared to the M4. The GPU also includes better graphics capabilities and 3rd-generation ray tracing, delivering up to 45% higher graphics performance than M4.  

The M5 has the world’s fastest performance core with a CPU that can have up to 10 cores, 6 for efficiency and up to 4 for performance. This combination gives up to 15% faster multithreaded performance over M4. The chip also includes an improved 16-core neural engine, a strong media engine, and nearly 30% more unified memory bandwidth reaching 153 GB/s.  

The M5’s efficient performance powers the new 14-inch MacBook Pro, iPad Pro, and Apple Vision Pro, helping each device perform at its best. All of these products are now available for pre-order.  

M5 ushers in the next major leap in AI performance for Apple’s silicon, said Johny Srouji, Apple’s senior vice president of Hardware Technologies. With the introduction of neural accelerators in the GPU, M5 delivers a significant boost to AI workloads and a substantial increase in graphics performance. The world’s fastest CPU core, a faster neural engine, and even higher unified memory bandwidth deliver far greater performance and capabilities to the MacBook Pro, iPad Pro, and Apple Vision Pro.  

A New GPU Architecture Designed for AI and Graphics 

With the next-generation GPU architecture in M5, every part of the chip is optimized for AI. Its 10-core GPU has a neural accelerator in each core, giving four times the peak GPU compute compared to M4 and over six times the AI performance. Versus M1 with M5, the new 14-inch MacBook Pro and iPad Pro can process AI-driven tasks much faster, such as running diffusion models in apps like Draw Things or using large language models locally. With platforms such as Web AI, M5 also delivers higher graphics performance, achieving up to 30% faster performance than M4 and up to 2.5x faster than M1. M5 also includes Apple’s third-generation Ray Tracing Engine, providing up to a 45% graphics uplift in apps. Using ray tracing combined with re-architected second-generation dynamic caching, the GPU provides improved gameplay, more realistic visuals in 3D applications, and faster generation times for complex graphics projects. And other visually intensive applications, with M5, Apple Vision Pro renders 10% more pixels than with its Micro-OLED displays, and refresh rates can reach up to 120Hz, resulting in crisper details, more fluid display performance, and reduced motion blur.  

The GPU architecture is designed for flawless integration with Apple’s software frameworks. Applications that use built-in Apple frameworks and APIs such as Core ML, Metal Performance Shaders, and Metal 4 can achieve immediate performance gains. Developers can also build solutions for their apps by directly programming the neural accelerators using Tensor APIs in Metal 4.  

A Faster Neural Engine for Smarter Features 

The faster 16-core neural engine delivers powerful AI performance with incredible energy efficiency, complementing the neural accelerators in the CPU and GPU to make M5 fully optimized for AI workloads. For example, AI-powered features on Apple Vision Pro, like the ability to transform 2D photos into spatial scenes in the Photos app or to generate a persona, operate at higher speeds and with greater capabilities.  

The Neural Engine in M5 also enhances performance for Apple’s on-device AI tools like Image Playground, which run faster, and the faster Neural Engine and Unified Memory in M5 improve the overall performance of Apple’s intelligence models. Also, developers using Apple’s Core ML framework will achieve faster performance.  

Improved Memory for More AI Capabilities 

The M5 has a unified memory bandwidth of 153 GB, which is almost 30% more than the M4 and over twice that of the M1. Unified memory lets the entire chip use a single, large pool of memory, so MacBook Pro, iPad Pro, and Apple Vision Pro can run larger AI models right on the device. It also powers the faster CPU/GPU and Neural Engine, giving better multi-threaded performance in apps, faster graphics in creative apps and games, and quicker AI performance.  

On the neural accelerators or neural engine with 32 GB of memory, M5 lets users run demanding creative programs like Adobe Photoshop and Final Cut Pro simultaneously, even while uploading large files to the cloud.  

Apple Silicon and the Environment 

Apple 2030 is the company’s ambitious plan to be carbon neutral across its entire footprint by the end of this decade, by reducing product emissions from its three biggest sources:  

1. Materials  

2. Electricity  

3. Transportation  

The power-efficient performance of M5 helps the new 14-inch MacBook Pro, iPad Pro, and Apple Vision Pro meet Apple’s high standards for energy efficiency and lowers the total energy used over each product’s lifetime.

Source: Apple unleashes M5, the next big leap in AI performance for Apple silicon