AI Model Release Tracker

GPT-6 Sol

Released

GPT-6 Sol is an AI model released by OpenAI on Tuesday, Sep 22 2026, 19 days after GPT-6 Astra. It has a 1.05M token context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider.. Benchmark results (shown below) cover Next.js Evals, CheatBench, BullshitBench v2, DeepSWE 1.1, Supabase Evals, Terminal-Bench 4.0, and 5 more.

API pricing

Up to 272K input tokens
InputWhat you pay for everything you send the model — your question, plus any documents or earlier conversation you include with it.
$2.00
Cached inputA reduced rate for text you send over and over. If every request starts with the same instructions or the same document, the provider keeps a copy ready and charges less to read it again.
$0.20
Cache writeA one-off charge for storing text so later requests can reuse it at the cheaper cached rate.
$2.50
OutputWhat you pay for the text the model writes back. It is normally the dearer half: producing an answer costs more than reading one.
$10.00
Over 272K input tokens
InputWhat you pay for everything you send the model — your question, plus any documents or earlier conversation you include with it.
$4.00
Cached inputA reduced rate for text you send over and over. If every request starts with the same instructions or the same document, the provider keeps a copy ready and charges less to read it again.
$0.40
Cache writeA one-off charge for storing text so later requests can reuse it at the cheaper cached rate.
$5.00
OutputWhat you pay for the text the model writes back. It is normally the dearer half: producing an answer costs more than reading one.
$15.00
  • Reasoning or thinking is supported.
  • OpenAI also publishes a Fast mode for this model at exactly twice the standard rate in every column — $4.00 per million input tokens and $20.00 per million output at short context.
USD per 1M tokensEvery price here is for one million tokens. A token is roughly three-quarters of a word, so a million tokens is about 750,000 words of text.Verified October 19, 2026Read from developers.openai.com. Click the check to open the source.

Available from

OpenAI—flex$1.00$5.001M
OpenAI——$2.00$10.001M
Amazon Bedrockglobal—$2.00$10.001M
Azure——$2.00$10.001M
Amazon Bedrockus-east-1—$2.20$11.001M
Azureeu—$2.20$11.001M
Azureus—$2.20$11.001M
OpenAI—fast$4.00$20.001M
USD per 1M tokensEvery price here is for one million tokens. A token is roughly three-quarters of a word, so a million tokens is about 750,000 words of text.Checked November 7, 2026Third-party rates from openrouter.ai, not the lab’s own list price.

Benchmarks

Coding

Next.js EvalsNext.js coding — Vercel's open eval of how well AI coding agents build and migrate real Next.js apps — measured as the share of tasks the agent completes successfully. Higher is better.
100%
#1
DeepSWE 1.1Agentic coding — Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. (Version 1.1 of the test.) Higher is better.
68.8%
#16 of 37
Supabase EvalsSupabase coding — Supabase's own open benchmark: a coding agent is dropped into a real Supabase project and asked to do real work — set up a schema, fix a broken security policy, debug an Edge Function — and every run is checked against a live Supabase stack. This is the headline number, where the agent has Supabase's own skills loaded, as most people building on Supabase would. The score is the share of scenarios it got right. Higher is better.
78.3%with skills
78.3%no skills
with skills
#8 of 11
no skills
#5 of 11

Terminal & CLI

Terminal-Bench 4.0Agentic terminal coding — Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Version 4.0 recalibrated how much time, CPU and memory each task gets, removed eight tasks and fixed nineteen, so fewer runs fail for reasons that have nothing to do with the model. Scores are not comparable with earlier versions. Higher is better.via Terminal-Bench
49.39%
#12 of 32

Agentic & tool use

OSWorld 2.0Agentic computer use — Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Version 2.0 is a harder, refreshed task set. Higher is better.
60.5%
#7 of 10
Agent's Last ExamAgentic computer use — A hard set of desktop and operating-system tasks an AI agent has to finish by looking at the screen and working the machine itself. The score is the share it passes outright — partial credit does not count. Higher is better.
56.4%pass@1
#3 of 11
AutomationBenchBusiness workflows — Tests whether the AI can run real multi-step business workflows — the kind of end-to-end office processes companies want to automate — from start to finish. Higher is better.
33.2%
#15 of 20

Reasoning & science

ARC-AGI-2Abstract reasoning — Puzzle-style tests of abstract reasoning and pattern-finding — the kind of thing people find easy but AIs often struggle with. Higher is better.via BenchLM
89.58%
#8 of 41

Robustness

CheatBenchCheating rate — Measures how often AI agents try to cheat on difficult assignments, such as reading hidden answers, copying work or manipulating grading. The overall score gives equal weight to ten categories; the sycophancy category measures how far an agent shifts its beliefs toward a user's stated views. Scores describe each model in its tested agent setup, and task success is measured separately. Lower is better.via CheatBench
71.9%
#8 of 13
BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better.
59%
#27 of 98

Community preference

threejsevalCommunity preference (Three.js) — Every model gets the same prompt — "the Eiffel Tower", "a glass fishbowl", "a robot arm picking toys into a box" — and builds a 3D scene in Three.js. Real people then see two scenes side by side, names hidden, and vote for the one they prefer. The votes become a chess-style Elo rating on threejseval.com, averaged across all the prompts. It measures whether the scene looks and moves right to a human eye, not whether the code passes a test. Higher is better.
1622
#10 of 29

Source: Terminal-Bench, retrieved 12 October 2026 · Source: BenchLM, retrieved 12 October 2026 · Source: CheatBench, retrieved 25 September 2026. Other sources are identified on the linked benchmark pages.

Compare GPT-6 Sol with

GPT-6 Sol

Suggested comparisons

GPT-6 SolvsGPT-6.1 TerraGPT-6 SolvsClaude Haiku 5.5GPT-6 SolvsGemini 4 CarbonGPT-6 SolvsMuse Spark OpenGPT-6 SolvsGrok 4.7GPT-6 SolvsDeepSeek-V4.1-Flash-1017GPT-6 SolvsMistral Large 4GPT-6 SolvsKimi K3.1 ProGPT-6 SolvsGLM-5.3-FlashGPT-6 SolvsQwen3.8-Max-0902GPT-6 SolvsNemotron 3.5 Bolt E4B 4GB

About

GPT-6 Sol was introduced in September 2026 alongside GPT-6 Luna as a lower-cost option within the family opened by Astra. OpenAI positioned Sol for coding, professional tasks and agent workflows, while reserving Astra for more demanding work. The initial product announcement offered the pair through ChatGPT Work, Codex and the API, with access dependent on the plan and interface.

The API model documented a 1,050,000-token context window, text and image inputs, and configurable reasoning effort. These controls allowed developers to adjust the computation used for different requests. Sol's launch emphasized efficiency, but comparisons with other models still depended on reasoning settings, tools and task selection. GPT-6.1 Sol followed as a separate update later that month.

Frequently asked questions

GPT-6 Sol was released by OpenAI on Tuesday, Sep 22 2026.

All OpenAI releases

49 tracked

2022

1 release
GPT-3.5
Nov 30 2022