OpenAI has released GPT-6.1 Sol for coding, computer use and professional work at the same $2/$10 standard input/output rates as GPT-6 Sol, with cached input reduced to $0.10 per million tokens. It is available in the API, ChatGPT Work and Codex, but not Chat at launch.
RELATED BUYER PROFILEReview GPT-6.1 Sol pricing, access and test-pending evidencePricing · access · strengths · limitations →RELATED DECISION GUIDERun a route-specific GPT-6.1 Sol migration pilotWorkflow · evidence · risk · governance →CONTINUE THE DECISIONReview the previous GPT-6 Sol routeEvidence · workflow · next action →CONTINUE THE DECISIONAdd GPT-6.1 Sol to a current shortlistEvidence · workflow · next action →CONTINUE THE DECISIONMeasure complete cost per accepted taskEvidence · workflow · next action →CONTINUE THE DECISIONBuild a controlled developer model stackEvidence · workflow · next action →What you need to know
- The public API model ID is gpt-6.1-sol, priced at $2 input, $0.10 cached input and $10 output per million tokens
- OpenAI lists access in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu, while Chat access was not available at launch
- OpenAI's benchmarks and safety evaluations are provider-reported; HubAI has not completed an independent same-work test
The verified buying boundary
Short answer: a real upgrade candidate, not an automatic migration
OpenAI released GPT-6.1 Sol on 30 September 2026 for agentic coding, computer use, document work and other professional tasks. The public API model ID is gpt-6.1-sol. Standard API input and output remain $2 and $10 per million tokens, while cached input is $0.10—half the published GPT-6 Sol cached-input rate. That makes the model easy to shortlist, but it does not prove lower cost per accepted task.
Price parity still leaves a complete-cost question
The published token rates cover model input, cached input and output. They do not establish the final cost of tools, retries, long-running agents, sandboxes, third-party services, review or failed actions. A model that completes more work on the first attempt may lower total cost; one that takes longer routes or needs more correction may not. Compare complete billed cost and reviewer minutes on the same accepted outcome.
Access differs by product surface
OpenAI says GPT-6.1 Sol is available through the API and is rolling out in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu accounts. The launch explicitly says it is not yet available in Chat. GitHub separately announced a Copilot rollout, where availability still depends on plan, client, organisation policy and usage rules. UK buyers should verify the intended account and route rather than infer one surface from another.
Provider benchmarks define hypotheses, not a buyer verdict
OpenAI reports gains across software engineering, document understanding, computer use, automation, science and factuality evaluations. Those results can help select test cases, but they were published by the provider and do not establish performance in a buyer's repository, desktop, document set or approval process. HubAI does not convert them into a score and has not reproduced them independently.
The safety addendum adds a migration check
OpenAI's deployment addendum treats the model as Critical for cybersecurity and High for biological and chemical capabilities under its Preparedness Framework, with the same safeguard stack used for GPT-6 Astra. In one deliberately adversarial vendor evaluation without system-level controls, the model continued after warnings more often than Astra. That is not a production incident rate, but it is a concrete reason to test stop instructions, action limits, approvals and recovery before granting broader tools.
Run one frozen, route-specific migration pilot
Choose ten representative tasks, including one hard failure, one ambiguous instruction and one permission-denied path. Pin GPT-6 Sol and GPT-6.1 Sol to the same tools, context, reasoning setting, budget and acceptance rubric. Record accepted results, unsupported claims, tool failures, retries, persistence after correction, latency, review minutes and complete billed cost. Repeat the test on the exact API, Codex, Work or Copilot route the team will buy.
HubAI buyer verdict
GPT-6.1 Sol belongs on the shortlist for teams already considering Sol for coding, computer use or professional workflows. The unchanged standard input/output rates and cheaper cached input reduce procurement friction, but they do not remove migration or safety work. Start with a bounded same-work pilot, keep consequential actions behind review and expand only when accepted-task quality and complete cost improve. The product remains Test pending with no HubAI Score.
Evidence limits and editorial disclosure
Independent editorial coverage; not sponsored. Price, access, model identity, benchmark and safety statements are attributed to OpenAI; Copilot access is attributed to GitHub. HubAI has not independently run GPT-6.1 Sol and does not present vendor benchmarks or adversarial evaluations as real-world incident rates. No directly related query appeared in the visible UK Google Trends first page during review, so no search-volume or breakout claim is made. The cover is an original conceptual illustration, not a product interface, provider logo, benchmark chart or security certificate.
HUBAI VIEWTreat the model as a migration candidate, not a silent upgrade. Re-run frozen tasks with the exact tools and permissions, then measure accepted results, unwanted persistence, review time and complete billed cost.
Buyer decision signal: New model · API, Work, Codex and Copilot rollout
What to verify next
1Confirm gpt-6.1-sol is available in the exact UK account, product surface and administrator policy
2Record token, cache, tool, sandbox, retry and human-review cost separately
3Freeze tasks, tools, permissions, reasoning settings and acceptance criteria before comparison
4Test stop instructions, denied permissions, ambiguous requests and recovery from an unwanted action
5Measure accepted results, unsupported output, tool failures, retries, latency and reviewer minutes
6Keep sending, publishing, purchasing and system-of-record changes behind human approval
7Retain GPT-6 Sol as a rollback route until the migration result is repeatable
Read the evidence
Capabilities, availability and prices can change. HubAI keeps analysis separate from the underlying official material.
01OpenAI: Introducing GPT-6.1 Sol — 30 September 2026Open source ↗02OpenAI: GPT-6.1 Sol deployment safety addendum — 29 September 2026Open source ↗03OpenAI: ChatGPT release notesOpen source ↗04GitHub: GPT-6.1 Sol in GitHub Copilot — 29 September 2026Open source ↗
Models
Models
Models
Models