Editorial illustration of a model inside a verification chamber for the GPT-6.1 Sol release

model-release

GPT-6.1 Sol Release Date: September 29, 2026 and Channel Availability

Q: Is GPT-6.1 Sol the same model as GPT-6 Sol?

No. The official identifiers are gpt-6.1-sol and gpt-6-sol. In particular, their supported reasoning effort and Chat Completions tool-calling contracts differ.

Jessie

COO

September 29, 2026

7 min read

GPT-6.1 Sol was released on September 29, 2026, at OpenAI DevDay. The official recap and model reference establish the release and public API identity. For teams already running GPT-6 Sol agents, the immediate question is how the new model changes their existing workflow: ordinary input and output list rates stay the same, cached reads cost less, and the tool-calling contract changes.

Start with the GPT-6.1 Sol vs GPT-6 Sol upgrade guide before changing an agent's model setting. EvoLink's GPT model collection provides the selection path across existing models. This article does not confirm GPT-6.1 Sol access or selling rates on EvoLink as of September 29. OpenAI availability and a verified gateway route are separate milestones.

Compare the configured model, token roles and integration choices on the GPT-6.1 Sol API page. A catalog listing or fallback price is not proof of a successful live request or a settled charge; verify those before routing production traffic.

GPT-6.1 Sol release date and availability by channel

Channel or milestone	Confirmed position on September 29, 2026	What to check next
Official release	Announced at DevDay on September 29	Use the version-qualified name, GPT-6.1 Sol
OpenAI API	Official reference lists `gpt-6.1-sol`	Account access, endpoint and supported controls
ChatGPT Work and Codex	The announcement lists Plus, Pro, Business, Enterprise and Edu	Your account and workspace's actual model access
ChatGPT Chat	Not yet available, according to the announcement	Do not infer Chat access from Work or Codex
Ultrafast in Codex	Announced for the coming days	A future offer is not a current speed guarantee
EvoLink	Access and rates are not verified in this article	Check the current model catalog and route documentation

The channel-specific claims above come from OpenAI's launch announcement. A subscription model picker is not evidence of an API rate or a gateway feature. Conversely, an official API listing does not guarantee that a particular client already recognizes the new identifier.

What is confirmed about the new Sol

GPT-6.1 Sol is a separate model from gpt-6-sol. The new reference documents the following contract:

Field	GPT-6.1 Sol
Official model ID and snapshot	`gpt-6.1-sol`
Input and output	Text and images in; text out
Context window	1,050,000 tokens
Maximum output	128,000 tokens
Knowledge cutoff	April 30, 2026
Reasoning effort	`low`, `medium` by default, `high`, `xhigh`, `max`
Unsupported effort levels	`none` and `minimal`
Tool calling	Use Responses; Chat Completions is supported without tool calling

The context window is a combined budget, not permission to send 1.05 million input tokens. Likewise, image input support does not mean that this model directly produces image or video output. Check the actual route's supported features before treating the provider reference as an integration contract.

The distinction matters most for old Sol agents. GPT-6 Sol permits Chat Completions function calling with reasoning_effort: none; 6.1 Sol removes that option. A model-name change alone cannot carry that workflow forward. The upgrade guide explains the endpoint and effort decisions without assuming that EvoLink already exposes the new route.

How to read the capability announcement

OpenAI's launch announcement reports a 6.4-percentage-point DeepSWE v1.1 improvement over GPT-6 Sol's best score, using lower effort and cost. This is a vendor-reported evaluation, not an equal-effort comparison or an EvoLink reproduction. The upgrade guide expands the evidence and its limits for coding, automation and computer use.

Use them to choose a trial, rather than to declare a winner. A coding team needs accepted patches on its repositories; a document team needs correct answers with traceable evidence; a business agent needs correct tool actions without duplicate writes. Different prompts, tools, reasoning effort and fallback policies can change both quality and task cost. This article reports the release, while the upgrade comparison turns those questions into an evaluation plan.

What the pricing change actually means

OpenAI's pricing page lists the same Standard ordinary input and output rates for 6 Sol and 6.1 Sol. The cache-read rate is lower on 6.1 Sol; cache writes remain a separate charge, and longer inputs have their own billing rules.

That creates a reason to evaluate agents that repeatedly reuse context. It does not establish a fixed discount on the whole job. A workload with little cache reuse can see little direct savings, while changes in reasoning output, retries or accepted-task rate can outweigh the cache-read difference. Keep the gateway's actual selling rates on its model page or pricing surface; do not substitute OpenAI's list rates for them.

What existing Sol teams should do next

Documents passing through evidence filters before a gate, illustrating GPT-6.1 Sol release and route verification

An editorial view of separating source evidence from claims that still need route verification; it is not a model-output sample.

Record your current model ID, endpoint, reasoning setting and tools first. Then decide whether the new provider contract can express the same workflow. If it can, compare a small, fixed set of representative tasks and include failures and reviewer time in the budget. If it cannot, estimate the integration work before evaluating performance.

For the current baseline, use the existing GPT-6 Sol model page. For broader model choice, use the GPT collection. The unified gateway selection path is useful for keeping a baseline selectable while evaluating a candidate, but availability, feature support and billing still need verification for each individual route.

FAQ

When was GPT-6.1 Sol released?

September 29, 2026, at OpenAI DevDay. The official recap and model reference distinguish it from GPT-6 Sol, released a week earlier.

Is GPT-6.1 Sol the same model as GPT-6 Sol?

No. The official identifiers are gpt-6.1-sol and gpt-6-sol. In particular, their supported reasoning effort and Chat Completions tool-calling contracts differ.

Is it available in ChatGPT Chat?

OpenAI's September 29 announcement says it is available in Work and Codex for the listed plans, but not yet in Chat. Check your account's actual access.

Does the OpenAI release confirm availability on EvoLink?

No. This article has not verified the new EvoLink route or selling rates. Check the current catalog and documentation before planning a call through the gateway.

Can I keep using Chat Completions for an agent with tools?

The official 6.1 Sol reference requires Responses for tool calling. Chat Completions remains supported without tools; the old none workaround is not supported on 6.1 Sol.

Should I replace GPT-6 Sol immediately?

Evaluate the new contract and accepted-task cost first. Retain the old baseline until compatibility, quality and billing checks pass. The linked upgrade guide provides the decision steps.

Sources

OpenAI DevDay 2026 recap — release event and date.
GPT-6.1 Sol launch announcement — channel availability and vendor-reported capability framing.
GPT-6.1 Sol model reference — exact identity, limits and request-contract boundaries.
GPT-6 Sol model reference — prior model's contract.
OpenAI API pricing — provider rates and processing conditions, checked September 29, 2026.

All Posts

#GPT-6.1 Sol #OpenAI #Release Date #Coding Agents