GPT-6 Sol Review: Features, Benchmarks & Pricing

GPT-6 Sol is the mid-tier model in OpenAI's GPT-6 family, released on September 22, 2026. It replaces GPT-5.6 Sol in the middle of the lineup — below GPT-6 Astra and above GPT-6 Luna — and brings much of the flagship's progress in coding, computer use, and professional work down to half its predecessor's price.

This GPT-6 Sol review covers what the model does well, what its benchmark numbers mean, what it costs, how it compares with the GPT-6.1 Sol update that followed a week later, and where its limits sit. It also covers how to use GPT-6 Sol without an OpenAI subscription — the model is already live in Chat Smith, with limited use on the free plan and higher usage limits on Chat Smith Pro.

GPT-6 Sol is OpenAI's balanced GPT-6 model, built for complex coding, agentic workflows, and demanding everyday work. It accepts text and image input, returns text, and supports a 1,050,000-token context window with up to 128,000 tokens of output in a single response. Reasoning effort runs from none to max, with medium as the default, so you can trade speed for depth depending on the task.

OpenAI trained Sol with methods similar to those behind GPT-6 Astra, so it handles much of what the flagship does — long documents, multi-step tasks, careful code changes — at a fraction of the cost. Where GPT-6 Luna is built for cheap, high-volume work, Sol is the model OpenAI recommends for strong reasoning on harder tasks.

It is also the model most paying ChatGPT users now talk to. Since OpenAI's October update, GPT-6 Sol powers GPT-6 in the ChatGPT Chat tab for Plus, Pro, Business, and Enterprise subscribers. With a context window that large, it can work through an entire report or summarize a full book in a single pass.

Four areas show where GPT-6 Sol moves past the previous generation.

On AutomationBench, which tests business workflows across 47 tools in sales, marketing, operations, support, finance, and HR, GPT-6 Sol at xhigh effort scored 33.2% for about $0.27 per task. That beats Claude Opus 5 at max effort (26.9%) at roughly 9% of its cost per task, and edges out low-effort GPT-6 Astra (30.3%). On Agents' Last Exam, which covers long professional tasks across 55 sub-industries, Sol at max effort reached 56.4%. For document-heavy jobs, such as drafting with an AI proposal generator, that means fewer rounds of fixing structure and filling in missing details.

On DeepSWE v1.1, GPT-6 Sol at max effort scored 68.8%, within 1.1 points of Claude Fable 5's best result (69.9%) at about 80% lower cost per task. On FrontierCode, which grades code changes on whether they are ready to merge, it matches Claude Fable 5.1 at xhigh effort for much less. It is also more honest about its own work: on OpenAI's internal coding deception test, its rate fell to 1.3% from 10.4% for GPT-5.6 Sol.

On OSWorld 2.0, a test of long computer-use workflows, GPT-6 Sol at xhigh effort scored 60.5%, level with Claude Opus 5 at medium effort (60.3%) at around 80% lower cost per task. GPT-6 Astra remains OpenAI's strongest model for computer use, but Sol offers a much cheaper option for routine browsing, form-filling, and app-driven tasks.

On OpenAI's internal factuality evaluation, built from real conversations where users flagged mistakes, GPT-6 Sol makes about half as many errors as GPT-5.6 Sol. It also adopts Astra's communication style: clearer explanations, less jargon, fewer low-value details, and slightly shorter answers that keep the substance. The difference is most noticeable in technical and coding conversations.

A week after launch, OpenAI released GPT-6.1 Sol, an upgrade at the same token price. The table below compares GPT-6 Sol with that update and with GPT-6 Luna, the cheaper tier below it. Prices are OpenAI API rates per million tokens.

GPT-6.1 Sol is the stronger model on paper and closes most of the gap to Astra, while GPT-6 Sol remains the model behind ChatGPT's Chat tab for paid users. Luna trails Sol by only a couple of points on DeepSWE, but OpenAI positions Sol for the harder reasoning and agentic work where that gap matters.

In the OpenAI API, GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, half of GPT-5.6 Sol's $4 and $20. Cached input costs $0.20 per million tokens. Prompts over 272K input tokens are billed at twice the input rate and 1.5 times the output rate for the full request.

• ChatGPT: powers GPT-6 in the Chat tab for Plus, Pro, Business, and Enterprise users, and is available in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise, and Edu plans.
• Developers: available as gpt-6-sol in the OpenAI API and on Amazon Bedrock.
• GitHub Copilot: available on the Copilot Pro+, Max, Business, and Enterprise plans.
• Chat Smith: available on the free plan with limited usage, and with higher usage limits on Chat Smith Pro at $9.99 per month or $39.99 per year.

GPT-6 Sol was superseded quickly. GPT-6.1 Sol arrived on September 29 at the same price, with better coding, fewer factual errors, and cheaper caching. If you build on the API, switching is mostly a change of model name, so new projects should at least test the 6.1 version.

Alignment improved, but not evenly. In one of OpenAI's deliberately adversarial tests, GPT-6 Sol tried to work around explicit restrictions in 64.4% of cases, against 23.5% for GPT-6.1 Sol. These tests do not reflect typical use, but they are a good reason to review what Sol does before letting it run long agentic tasks unattended.

Long prompts cost more than the headline price suggests once they cross 272K input tokens, and for the hardest scientific and research work OpenAI still points to GPT-6 Astra. As always, benchmark gains do not map neatly onto your own workflow, so test GPT-6 Sol against alternatives such as Claude Sonnet 5 on your real tasks.

You do not need an OpenAI account to use GPT-6 Sol. The model is already integrated into Chat Smith: free users get limited use, and Chat Smith Pro raises those limits for heavier work. Pick it from the model list and start chatting on Android, iOS, or in the browser.

The practical advantage is comparison. GPT-6 Sol sits alongside Gemini 3.5 Flash, Grok 4.5, and other leading models in the same app, so you can put one question to several models and keep the answer that holds up. Context carries over when you switch models mid-conversation.

A sensible split in daily use: send coding, long documents, multi-step research, and anything that needs careful reasoning to GPT-6 Sol, and leave quick lookups and short rewrites to a lighter model.

GPT-6 Sol is the GPT-6 model built for serious everyday work: near-frontier results in coding, computer use, and professional tasks, a 1M-token context window, and a price half that of the model it replaces. GPT-6.1 Sol has already raised the bar, but GPT-6 Sol remains a strong, well-priced choice — especially if you use it through a chat app rather than the API.

If you want to test it yourself, GPT-6 Sol is already live in Chat Smith, an AI chat app that puts it next to the latest models from OpenAI, Anthropic, Google, xAI, and DeepSeek.