GPT-6.1 Sol Explained: Astra-Level Coding at a Fifth of the Cost
By ChatGBot · September 30, 2026
At its DevDay on September 29, 2026, OpenAI unveiled GPT-6.1 Sol, an upgrade to GPT-6 Sol that nearly matches the top-tier Astra model on agentic coding, computer use, and professional work, but at roughly a fifth of the cost. It is aimed squarely at developers and agent builders, and it keeps the September price war going. Here is what changed and whether it matters for you.
Every top AI in one app
GPT, Claude, DeepSeek, Qwen and GLM — one subscription.
Try ChatGBot →Astra performance, a fraction of the price
The headline is value. On the DeepSWE v1.1 coding benchmark, GPT-6.1 Sol matches GPT-6 Astra at about one-fifth of the cost. It also improves on GPT-6 Sol across the board: 6.4 points higher on DeepSWE at lower reasoning effort, 4.8 points higher on AutomationBench at medium effort, and more than double the score on Terminal-Bench Science at maximum effort for less than half the cost per task. It still trails Astra on the very hardest tasks, but for everyday coding and agent work the gap is small and the savings are large. For context on the family, see our GPT-6 Sol vs Luna guide.
A very large context window
GPT-6.1 Sol takes a 1,050,000-token input window and can produce up to 128,000 tokens of output, which is enough to hold large codebases or long documents in a single request. One thing to watch: prompts over 272,000 tokens are billed at 2x input and 1.5x output for the whole request, so very long contexts cost more.
Pricing
In the API, GPT-6.1 Sol is $2 per million input tokens and $10 per million output tokens, with cached input at just $0.10. That is the same headline price as GPT-6 Sol, but you get materially better results per dollar. New to the family? Start with what is GPT-4o.
Built for agents and coding, not chat
GPT-6.1 Sol launched in ChatGPT Work and Codex rather than the regular Chat product, and it is available through the OpenAI API as gpt-6.1-sol, plus OpenRouter, Vercel AI Gateway, and GitHub Copilot. Reasoning is always on, with effort levels from low up to max. The Agents API also gained computer use and Codex-style multi-agent features, tool search, tool calling, and context compaction, which is where this model is meant to shine.
How it fits the bigger picture
GPT-6.1 Sol lands a week after Claude Opus 5.5 and the first GPT-6 Sol and Luna tiers, and the theme is the same: strong models are getting cheaper, and the advantage goes to whoever matches the right model to each task. For the full month in one place, read our September 2026 AI model releases roundup, and our take on Claude Opus 5.5.
Frequently asked questions
When was GPT-6.1 Sol released? September 29, 2026, announced at OpenAI's DevDay.
How much does GPT-6.1 Sol cost? $2 per million input tokens and $10 per million output tokens in the API, with cached input at $0.10.
Is GPT-6.1 Sol as good as GPT-6 Astra? It matches Astra on agentic coding benchmarks like DeepSWE v1.1 at about a fifth of the cost, but still trails Astra on the hardest tasks.
Where can I use GPT-6.1 Sol? In ChatGPT Work and Codex, and through the OpenAI API as gpt-6.1-sol, plus OpenRouter, Vercel AI Gateway, and GitHub Copilot.
Can I use GPT in ChatGBot? Yes. ChatGBot gives you GPT along with Claude, DeepSeek, Qwen and GLM under one subscription, so you can pick the right model for each task.