Positioning and release
GPT-5.6 Sol is OpenAI’s flagship GPT-5.6 model for complex professional work. OpenAI announced a limited preview on June 26, 2026, initially available to selected trusted partners through the API and Codex. On July 9, 2026, OpenAI announced broader availability across ChatGPT, Codex and the API. (openai.com)
Capabilities and uses
- Supports text input and output and image input. The API documentation lists a 1,050,000-token context window, a 128,000-token maximum output and a knowledge cutoff of February 16, 2026. (developers.openai.com)
- Designed for demanding coding, research and agentic workflows. OpenAI reports improvements in command-line engineering, quantitative biology and cybersecurity evaluations. It describes legitimate uses including code review, vulnerability research, patch development and defensive testing. (openai.com)
- API reasoning effort options are
none,low,medium,high,xhighandmax;mediumis the default. Streaming, function calling and structured outputs are supported. The Responses API supports tools including web search, file search, code interpreter and computer use. (developers.openai.com)
Limitations and safeguards
The model does not directly support audio or video, and fine-tuning is not supported. Access and rate limits depend on the service and usage tier. OpenAI applies safeguards to sensitive cyber and biological requests and prohibited uses. Its preview evaluation did not establish that the model could autonomously produce a complete exploit chain under the tested conditions; results should not be generalized to every real-world configuration. (developers.openai.com)
OpenAI API pricing
Checked November 16, 2026. The official model page lists promotional text-token prices in USD per 1 million tokens: $4.00 input, $0.40 cached input and $20.00 output. It states that the promotion is available until November 30, 2026. Cache writes are billed at 1.25 times the uncached input rate. For prompts exceeding 272,000 input tokens, the entire request is billed at 2 times the input rate and 1.5 times the output rate. Tool-specific charges may also apply. (developers.openai.com)
These are OpenAI API prices, not ChatGPT subscription or third-party cloud prices. OpenAI’s original preview announcement listed $5 input / $30 output per 1 million tokens; those are historical launch prices, not the promotional prices above. Regional-processing and FedRAMP endpoints may incur a 10% uplift under OpenAI’s pricing rules. (openai.com)