GPT-6.1 Sol vs GPT-6 Sol: Should You Switch?

OpenAI released GPT-6.1 Sol on September 29, 2026, one week after GPT-6 Sol. It is an upgrade to the same mid-tier model rather than a new tier: the standard price stays at $2 per million input tokens and $10 per million output tokens, cached input drops from $0.20 to $0.10 per million, and OpenAI reports clear gains on coding, computer use and multi-step business work. GPT-6 Sol remains available at its unchanged price.

This guide answers one question: should you move a GPT-6 Sol workload to GPT-6.1 Sol now? If you are instead choosing between Sol and GPT-6 Astra, see the Astra comparison linked below.

The short answer

For most API and Codex workloads, switch after a short regression run. GPT-6.1 Sol costs the same or less for every request, cuts cached input in half, and OpenAI reports it ahead of GPT-6 Sol on every launch evaluation where it compares the two. Stay on GPT-6 Sol for now if your prompts and evaluations are pinned for compliance or change control, if GPT-6.1 Sol has not reached your ChatGPT workspace yet, or if the tool you use does not list GPT-6.1 Sol.

What stays the same and what changes

Every price except cached input is identical, so the upgrade never raises a bill at the same token counts. The saving grows with how much of each request is served from cache.

What OpenAI reports GPT-6.1 Sol improves

OpenAI's launch post compares GPT-6.1 Sol with GPT-6 Sol on several evaluations. These are OpenAI's own results, run in its research environment or through the API, not independent measurements:

• DeepSWE v1.1 (software engineering in real codebases): 6.4 percentage points above GPT-6 Sol's best result, at a lower reasoning effort and cost; OpenAI says it matches GPT-6 Astra at roughly one-fifth of the cost.
• AutomationBench 1.0.6 (multi-step business workflows): up 4.8 points from GPT-6 Sol at medium reasoning effort.
• OSWorld 2.0, offline set (computer use): 7 points above GPT-6 Sol at maximum effort for less than half the cost.
• Terminal-Bench Science 0.1 (scientific workflows): more than double GPT-6 Sol's result at maximum effort, at less than half the cost per task.
• Factuality on difficult prompts: at low reasoning effort, the share of answers with a factual error falls from 11.4% to 7.7%.

OpenAI also reports better alignment results than GPT-6 Sol, such as disclosing a broken search tool more often (failing to disclose in 2.1% of cases against 4.9%). Treat all of these as reasons to test, then confirm on your own tasks at the reasoning effort you actually ship.

What the switch costs or saves

Because only cached input changes, the saving depends on your cache hit rate. These examples use standard short-context API rates.

In Codex the same pattern holds in credits: the agent turn above uses about 5.3 credits on GPT-6 Sol and 4.9 on GPT-6.1 Sol. Workloads with little cache reuse see no price change, so their case for switching rests on the capability gains alone.

Where each model is available

GPT-6.1 Sol is available in the OpenAI API as gpt-6.1-sol. In ChatGPT, OpenAI is rolling it out in ChatGPT Work and Codex starting with Pro and expanding to Plus, Business, Enterprise and Edu; Enterprise and Edu workspace owners enable it through model permissions. It is not yet available in Chat. GPT-6 Sol stays available in the API, ChatGPT Work and Codex.

Outside OpenAI, GitHub made GPT-6.1 Sol generally available in Copilot on September 29 for Copilot Pro+, Max, Business and Enterprise, billed at list price. Other tools add models on their own schedules, so check each vendor's model list before assuming the new version is there.

How to switch safely

1. Run your existing evaluation set on gpt-6.1-sol at the reasoning effort you use today, and compare accepted results, retries and total tokens rather than a single benchmark.
2. Keep prompts and tool definitions unchanged for the first run so any difference comes from the model.
3. Check cache hit rates after switching; the upgrade only saves money where requests reuse a stable prompt prefix.
4. Update model IDs in configuration, workspace defaults and automations deliberately. OpenAI has not said that GPT-6.1 Sol replaces GPT-6 Sol as a default anywhere, so do not assume an alias moved.
5. Keep GPT-6 Sol as a fallback until the new model has run cleanly in production for a representative period.

Bottom line

GPT-6.1 Sol is a same-price upgrade with cheaper cache reads and OpenAI-reported gains across coding, computer use and business workflows, so most teams should plan to move. The reasons to wait are operational: pinned evaluations, a workspace that has not received the rollout, or a tool that does not list the new model yet.