GPT-6 Sol Explained: Pricing, 1M Context Window, Limits, and Who Should Use It

OpenAI has released GPT-6 Sol, the mid-priced model in its new GPT-6 family, and it's built for people who need serious coding and agent performance without flagship prices. It went live on September 22, 2026, at $2 per million input tokens and $10 per million output tokens, with a context window of about 1.05 million tokens.

That's a strong headline, but the details matter here. The big context window comes with a pricing rule that catches people out, and there are a few practical limits worth knowing before you build on it. Here's the full picture.

GPT-6 Sol at a Glance

• Released: September 22, 2026
• Standard price: $2 per million input tokens, $10 per million output tokens
• Context window: 1,050,000 tokens, with up to 128,000 tokens of output
• Input and output: text and images in, text out
• Reasoning levels: none, low, medium (default), high, xhigh, and max
• Knowledge cutoff: April 20, 2026

What GPT-6 Sol Is Built For

OpenAI describes Sol as a model for complex coding and agentic workflows, meaning tasks where the AI plans, uses tools, and works through many steps on its own. It supports function calling, structured outputs, web search, file search, code execution, and computer use through the Responses API, which covers most of what teams building AI agents ask for.

It sits in the middle of the GPT-6 range. The flagship, GPT-6 Astra, arrived in early September, and Sol followed as the more affordable option for everyday professional work. The reasoning control is a nice touch: you can turn reasoning off entirely for quick, simple requests, or push it up to the maximum setting for hard problems.

The Pricing Details That Actually Matter

At $2 and $10 per million tokens, Sol is a clear step down from the $5 input and $30 output OpenAI listed for GPT-5.6 Sol when that generation launched in July. There are also several ways to bring the bill down further. Cached input costs just $0.20 per million tokens, and both Batch and Flex processing run at half the standard rates.

Now the catch. Prompts longer than 272,000 input tokens are billed at double the input rate and one and a half times the output rate, and that applies to the entire request, not only the tokens above the line. So while the 1.05 million token window is available, filling it gets expensive quickly. If you plan to feed Sol very long documents, budget around that threshold rather than the headline price.

Limits Worth Knowing Before You Build

A few practical points from OpenAI's documentation. Sol is not available on the free API tier, and starting-tier accounts get 500 requests and 500,000 tokens per minute. Fine-tuning isn't supported. Audio and video inputs aren't supported either, since the model accepts text and images only. And if you use the older Chat Completions endpoint, function calling only works with reasoning turned off, so agent-style tools are best run through the Responses API.

How It Compares

Independent benchmarking is still catching up, and the company's own claims about accuracy are best treated as just that until more third-party tests land. If you're weighing your options, it's worth lining Sol up against other recent releases, including Anthropic's Claude Opus 5.5, which launched the same day, and Alibaba's Qwen3.8-Max, which we covered last month. Price, speed, and context handling differ enough between them that the right pick depends on your workload.

What This Means for Anyone Working With Files

Because that pricing jump kicks in at 272K tokens, keeping your inputs lean pays off directly. A bloated PDF or a poorly scanned document burns through tokens without adding anything useful. Compressing a file, pulling out only the pages you need, or converting it to clean editable text before it goes anywhere near an API keeps both quality up and cost down. Our PDF Compressor, Split PDF, and PDF to Word tools handle that prep in your browser.

Developers will also be handling a lot of structured data, since Sol supports JSON schema outputs. A JSON Beautifier makes payloads easier to inspect, and a CSV to JSON converter gets spreadsheet data ready for an API call.

Who Should Try GPT-6 Sol

Teams building coding assistants, workflow automation, or multi-step agents on the OpenAI API have the most to gain, especially if they were previously running a pricier model for tasks that don't need a flagship. If you're only using ChatGPT day to day, check OpenAI's release notes for which model your plan uses, since the API and the app don't always update on the same schedule.

We'll keep following GPT-6 Sol as independent benchmarks arrive, in our AI tools and updates section.