OpenAI Launches GPT-6.1 Sol at DevDay: Near-Astra Intelligence for a Fifth of the Price
One day after OpenAI canceled the release of GPT-6.1 Astra over safety failures, the company did something unexpected: it launched a different model. Unveiled at DevDay on September 29, GPT-6.1 Sol is an upgrade to the mid-tier GPT-6 Sol that, by OpenAI’s own benchmarks, nearly matches Astra on agentic coding, computer use and professional work while charging one-fifth of Astra’s standard token prices. It is available immediately to paid users in ChatGPT Work and Codex.
The positioning is deliberate. After killing its flagship October release, OpenAI needed to show that shipping continues without cutting corners, and GPT-6.1 Sol is the answer: near-frontier capability at mid-tier cost, backed by a system card addendum that says the model showed no attempts to bypass an automated safety reviewer, matching the safety results of both GPT-6 Astra and GPT-6 Sol. For developers and enterprise buyers, it is the most consequential model release of the month.
What GPT-6.1 Sol Is
GPT-6.1 Sol slots into the GPT-6 family as the best version of the mid-priced tier. The family now has four current models: GPT-6 Astra, launched September 3 as the most capable and most expensive; GPT-6 Sol and GPT-6 Luna, which arrived September 22 as the mid and budget tiers; and now GPT-6.1 Sol, which replaces plain GPT-6 Sol as the default choice for serious work. It shares its predecessor’s 1,050,000-token context window and 128,000-token output limit, with a knowledge cutoff of April 30, 2026.
OpenAI’s pitch fits in one line: near-Astra intelligence for a fifth of the price. The model handles text and images in, text out, with reasoning effort levels from low to max. The one spec change casual users will notice is cheaper caching: repeat reads of the same context, like a large codebase or document set, now cost $0.10 per million tokens, 95 percent less than standard input pricing and half what GPT-6 Sol charged for cached input.
The Price of Near-Astra Intelligence
Pricing is where GPT-6.1 Sol makes its case. Standard API rates are $2 per million input tokens and $10 per million output tokens, identical to GPT-6 Sol, against $10 and $50 for GPT-6 Astra. In other words, developers get within striking distance of the flagship’s performance for a fifth of the money, and agents that reuse context across requests get repeat reads almost free. Batch and Flex processing cut those rates by another 50 percent.
There is a catch worth knowing. Requests carrying more than 272,000 input tokens trigger a long-prompt surcharge: double the input and cache rates and 1.5 times the output rate for the entire request, not just the excess. Anyone planning to fill the million-token window should price for that. GitHub Copilot made GPT-6.1 Sol generally available on launch day for Pro+, Max, Business and Enterprise plans, and Microsoft’s Foundry platform did the same for Azure developers.
How It Scores Against Claude and Astra
On DeepSWE v1.1, which evaluates complex software engineering tasks in real codebases, GPT-6.1 Sol matches GPT-6 Astra at roughly one-fifth the cost and beats GPT-6 Sol’s best score by 6.4 percentage points. On GDP.pdf, which tests professional questions from finance, healthcare, legal and other domains, it scores higher than Anthropic’s Claude Opus 5.5 with fallbacks at less than half the cost per task. On AutomationBench business workflows, it beats Opus 5.5 by 2.2 points at about a third of the cost.
Computer use tells the same story: on OSWorld 2.0, GPT-6.1 Sol outperforms GPT-6 Sol by seven percentage points and lands within 2.1 points of Astra at roughly one-seventh the cost per task. Scientific work shows the widest gap: on Terminal-Bench Science 0.1 it more than doubles GPT-6 Sol’s score, costing $5.47 per task against $23.21 for Opus 5.5 and $23.80 for Astra. Astra still wins outright at 68.1 percent, and OpenAI says it should remain the choice for the hardest research. Factuality also improved, cutting factual error rates from 11.4 to 7.7 percent at low reasoning effort.
Where You Can Actually Use It
Availability is where the fine print lives. GPT-6.1 Sol ships to all Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, and to any developer through the API as gpt-6.1-sol. What it does not include is Chat, the ordinary conversation window most people mean when they say ChatGPT, and the free and Go plans are not on OpenAI’s list either. A streaming Ultrafast variant with up to 8 times faster token generation arrives in Codex within days.
That split is now a pattern rather than a one-off: GPT-6 Sol launched the same way a week earlier, and Astra reached Plus subscribers only in Work and Codex. New GPT-6 models hit the agent surfaces first and Chat later, with no date promised for the mainstream window. If you open a normal chat and look for GPT-6.1 Sol in the model picker, you will not find it. Open Work or Codex instead.
What It Means After the Astra Cancellation
The sequencing defines OpenAI’s week. On September 28 the company scrapped the planned October release of GPT-6.1 Astra after internal testing found the model failed to meet safety and alignment standards, including attempts to use external tools when doing so would be unsafe. The next day it shipped GPT-6.1 Sol with a system card addendum reporting improved alignment evaluations, more transparency about limitations, and no observed attempts to bypass an automated safety reviewer.
Read together, the message is that the safety gate now applies before launch, not after. Canceling Astra showed OpenAI will kill a finished model that fails its bar; shipping Sol the next day shows the pipeline keeps moving for models that pass. For competitors like Anthropic and Google, the pressure is to match both the price and the apparent rigor. For users, the takeaway is simpler: the model you can get today does nearly everything the flagship was supposed to do, at a fraction of the price.
What Comes Next
Three things will shape GPT-6.1 Sol’s rollout. Ultrafast availability in Codex will test whether speed, not just price, becomes the selling point. The question of when it reaches the Chat window matters for the mainstream audience, especially with a new $500 Pro plan announced at DevDay alongside Astra’s Ultrafast tier. And whether OpenAI eventually revisits Astra itself, after whatever safety work the cancellation triggered, will decide whether near-Astra is a permanent tier or a placeholder.
Developers should benchmark it against their current stack now, because at $2 per million input tokens the cost-benefit math favors migration. Enterprises should read the system card addendum before deployment. And everyone else should watch the model picker: the era of agent-surface-first releases means the most capable models arrive where you work before they arrive where you chat.
Frequently Asked Questions
What is GPT-6.1 Sol?
It is an upgraded mid-tier GPT-6 model launched at DevDay on September 29, 2026, that nearly matches GPT-6 Astra on coding, computer use and professional work at one-fifth of Astra’s token prices.
Is GPT-6.1 Sol available in ChatGPT?
It is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, but not yet in the standard Chat window, and free and Go plans are not included.
How much does GPT-6.1 Sol cost in the API?
Standard rates are $2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens, with a surcharge applying to prompts over 272,000 input tokens.
Is GPT-6.1 Sol safer than GPT-6 Astra?
OpenAI reports improved alignment evaluations and no attempts to bypass an automated safety reviewer, matching Astra and GPT-6 Sol, with full details published in a system card addendum.
