GPT-6 Sol is OpenAI’s balanced GPT-6 model for interactive coding, agentic development and complex production work that does not require the higher-priced GPT-6 Astra. The API model ID is gpt-6-sol. It accepts text and images, returns text, and has a 1,050,000-token context window with up to 128,000 output tokens.
OpenAI positions Sol between Astra and GPT-6 Luna: use Astra for the hardest frontier work, Sol for demanding general engineering, and Luna for focused high-volume tasks. We have not independently benchmarked these models.
GPT-6 Sol pricing
Prices are per million tokens.
| Processing | Input | Cached input | Cache write | Output |
|---|---|---|---|---|
| Standard | $2.00 | $0.20 | $2.50 | $10.00 |
| Batch / Flex | $1.00 | $0.10 | $1.25 | $5.00 |
| Fast | $4.00 | $0.40 | $5.00 | $20.00 |
Batch and Flex share a price but not an operating model: Batch is asynchronous, while Flex exchanges stronger availability and latency guarantees for lower cost. Fast is the premium low-latency route.
For requests above 272,000 input tokens, OpenAI applies the higher rate to the whole request: $4/M input, $0.40/M cached input, $5/M cache writes and $15/M output. Standard processing is the only option documented with EU data residency; regional processing adds 10%, and Fast does not support EU data residency.
Capabilities and API boundaries
Sol supports the Responses API, Chat Completions and Batch. OpenAI recommends Responses for agentic tool use. Chat Completions supports function calling only when reasoning is set to none, so do not assume an older chat integration exposes the full agent feature set.
Reasoning levels are none, low, medium, high, xhigh and max, with medium as the default. Higher effort can increase latency and token consumption; select it through workload evaluations rather than globally.
Sol in GitHub Copilot
GitHub announced a gradual rollout of Sol to Copilot Pro+, Max, Business and Enterprise on September 22, 2026. Supported pickers include VS Code, Visual Studio, Copilot CLI, the coding agent, Copilot App, github.com, GitHub Mobile, JetBrains, Xcode and Eclipse. Business and Enterprise administrators can manage access through model policies.
Copilot availability is separate from direct OpenAI API access and billing. Copilot uses usage-based billing for the selected model.
When to choose Sol
Choose Sol for repository-scale coding, multi-step tool workflows and engineering analysis when Luna misses your quality target but Astra’s premium is not justified. Measure successful-task cost, retries, tool accuracy and review time—not only token prices.
For simpler classification, extraction and short code changes, Luna can be twenty times cheaper on standard input and output pricing. For the hardest autonomous work, compare Sol with Astra on a representative evaluation set through our AI Testing & Evaluation hub.
Verdict
GPT-6 Sol is the practical GPT-6 default for serious development work: substantially cheaper than Astra while retaining the same 1.05M context and 128K output limits. Long-context multipliers and processing modes can materially change the bill, so model those before migrating a large-context agent.
Compare GPT-6 Sol vs Claude Opus 5.5 for the cross-provider decision, or GPT-6 Sol vs GPT-6 Luna for the same-family cost and capability tradeoff.