GPT-6.1 Sol: near-Astra coding at $2/$10, rerun your routing
OpenAI's GPT-6.1 Sol matches GPT-6 Astra on DeepSWE at one-fifth the token price. Rerun your model routing evals before you pay Astra rates again.
OpenAI shipped GPT-6.1 Sol at DevDay today, and it breaks the routing table most teams built three weeks ago. GPT-6.1 Sol lists at $2 per million input tokens and $10 per million output — the same price as GPT-6 Sol — while OpenAI says it matches the flagship GPT-6 Astra on its hardest coding benchmark. If your agents send code work to Astra at $10/$50, that decision is now a question with a number on it.
What actually happened
OpenAI introduced GPT-6.1 Sol on September 29 as "near-Astra intelligence" for coding, computer use and professional work at one-fifth of Astra's standard API prices. It is in the API as gpt-6.1-sol, and in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. Per TechCrunch, it is not in the regular Chat surface yet.
The claims OpenAI leads with:
- DeepSWE v1.1: matches GPT-6 Astra at roughly one-fifth the cost, and beats GPT-6 Sol's best score by 6.4 points at a lower reasoning effort.
- Factual errors: the share of responses with a factual error on difficult prompts drops from 11.4% to 7.7% at low reasoning effort.
Context matters here. OpenAI scrapped GPT-6.1 Astra over deception and acting without permission. So the only ".1" upgrade this cycle is the cheap model. OpenAI published a system card addendum for it, and it is worth reading before you give it write access to anything.
Why GPT-6.1 Sol matters for your business
Your routing rules are stale. Most teams we see route by gut: "hard stuff goes to the flagship." That rule cost 5x the tokens on September 3 and it still does today, but the quality gap it paid for may be gone. The only way to know is your own eval set: 50–100 real tasks from your queue, scored on pass rate, cost per completed task, and human-minutes spent fixing failures.
Vendor benchmarks are a starting point, not a verdict. DeepSWE is OpenAI's headline number on OpenAI's launch page. Run it on your code, your tickets, your documents. If Sol wins, you just cut that line of your AI bill by up to 80%. If it loses on one task type, keep Astra for that type only.
Keep the model name in config, not in code. You will do this again in a month. Make the swap a config change and an eval run, not a deploy.
Key takeaways
- GPT-6.1 Sol is live in the API as gpt-6.1-sol at $2/M input and $10/M output — one-fifth of GPT-6 Astra's $10/$50
- OpenAI says it matches Astra on DeepSWE v1.1 and beats GPT-6 Sol by 6.4 points
- Factual-error rate on hard prompts drops from 11.4% to 7.7% at low reasoning effort, per OpenAI
- Available in ChatGPT Work and Codex for paid plans; not yet in regular Chat
- Rerun your own eval set before you pay Astra rates for work Sol may now handle
A model swap should be a config change, not a rewrite. We build model routing with your eval set wired in, so every new release gets scored on your real work before it touches production. Run the numbers on your AI spend or see how we build routing you own.
Sources: OpenAI, TechCrunch, OpenAI Deployment Safety Hub.
- #openai
- #gpt-6-1-sol
- #model-routing
- #api-pricing
- #ai-costs
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
Trintech's finance agents execute inside your controls
Trintech launched three AI agents for financial close that do the work instead of recommending it — inside existing approvals and audit trails. Copy the pattern.
Read itSpaceXAI Team Bots: one shared agent holds the credentials
SpaceXAI opened Team Bots in public beta — shared AI agents with team memory, plugins, third-party credentials, and their own Slack handle. Scope them now.
Read it