GPT-6.1 Sol is OpenAI’s current balanced GPT-6 model for coding, agents, and demanding production work. Its API model ID is gpt-6.1-sol. It accepts text and images, returns text, and provides a 1,050,000-token context window with up to 128,000 output tokens.

This page previously covered gpt-6-sol. That earlier model remains a separate version with its own API record. Do not silently substitute one model ID for the other. GPT-6.1 Sol keeps the same $2/M input and $10/M output headline rates, but lowers cached input from $0.20/M to $0.10/M and changes documented reasoning and tool boundaries.

OpenAI positions Sol below GPT-6 Astra and above GPT-6 Luna. Use Astra for the hardest frontier tasks, Sol for balanced engineering work, and Luna for focused high-volume work. We have not independently benchmarked GPT-6.1 Sol.

GPT-6.1 Sol pricing

Prices are per million tokens.

Processing	Input	Cached input	Cache write	Output
Standard	$2.00	$0.10	$2.50	$10.00
Batch / Flex	$1.00	$0.05	$1.25	$5.00
Fast	$4.00	$0.20	$5.00	$20.00

Batch and Flex share discounted rates but serve different workloads. Batch is asynchronous. Flex suits work that can tolerate weaker availability and latency guarantees. Fast is the premium low-latency route.

For requests above 272,000 input tokens, OpenAI applies higher rates to the whole request. Input, cached input, and cache writes become 2 times their standard rate, while output becomes 1.5 times its standard rate. That means $4/M input, $0.20/M cached input, $5/M cache writes, and $15/M output.

Standard processing supports US and EU data residency. Fast is not available with EU data residency. Teams with regional requirements should verify the selected processing mode rather than assume model availability implies regional availability.

Capabilities and API boundaries

GPT-6.1 Sol supports the Responses API, Chat Completions, and Batch. OpenAI recommends Responses for tool-using agents. Chat Completions is available without tool calling, so an existing chat integration is not equivalent to a Responses-based agent.

Documented reasoning levels are low, medium, high, xhigh, and max, with medium as the default. Unlike the earlier GPT-6 Sol record, GPT-6.1 Sol does not document none or minimal reasoning. Review code that depends on a no-reasoning mode before migrating.

The Responses API tool set includes functions, web search, file search, computer use, image generation, code interpreter, hosted shell, apply patch, skills, MCP, and tool search. Availability can still depend on endpoint, account, and tool configuration. The OpenAI Agents API comparison explains when a managed agent harness is more appropriate than direct model and tool calls.

GPT-6.1 Sol in GitHub Copilot

GitHub announced a gradual rollout of GPT-6.1 Sol to Copilot Pro+, Max, Business, and Enterprise on September 29, 2026. GitHub lists VS Code, Visual Studio, Copilot CLI, coding agent, Copilot App, github.com, GitHub Mobile, JetBrains, Xcode, and Eclipse as supported surfaces. Business and Enterprise administrators manage access through model policies.

Copilot availability is separate from direct OpenAI API entitlement and billing. A model appearing in Copilot does not prove that the same account can call its OpenAI API model ID.

Legacy GPT-5 migration deadline

OpenAI deprecated gpt-5.1 and gpt-5.3-codex on October 1, with shutdown scheduled for April 1, 2027. Both remain callable until then. Its deprecation table names gpt-6-sol, not gpt-6.1-sol, as the migration target. Do not silently equate the versions; validate any further move to 6.1 separately.

Migrating from GPT-6 Sol

Treat the move from gpt-6-sol to gpt-6.1-sol as a versioned migration:

1. Change the model ID explicitly in a test environment.
2. Remove assumptions about none reasoning and Chat Completions tool calling.
3. Recalculate cache-heavy workloads using the lower $0.10/M cached-input rate.
4. Re-run tool-use, long-context, latency, and output-quality evaluations.
5. Keep rollback and model-version telemetry until the new version is stable.

The version change does not remove the need for workload-specific testing. Cached-input savings can matter for repeated agent context, but prompts above 272K still trigger the whole-request multiplier.

When to choose GPT-6.1 Sol

Choose GPT-6.1 Sol for repository-scale coding, multi-step tool workflows, and engineering analysis when Luna misses the quality target but Astra’s premium is not justified. Measure successful-task cost, retries, tool accuracy, review time, and latency rather than token prices alone.

For simpler classification, extraction, and short code changes, Luna remains far cheaper. For high-stakes autonomous work, compare Sol with Astra on a representative evaluation set through our AI Testing and Evaluation hub. For production architecture, use the AI Application Architecture hub and AI Operations hub.

My take

GPT-6.1 Sol is the practical current default for teams that want the GPT-6 tool ecosystem without Astra pricing. The lower cached-input rate improves repeated-context economics, but the version is not a drop-in rename. The loss of none reasoning and the Chat Completions tool boundary are migration details that can break an integration even when headline prices look familiar.

Keep the old GPT-6 Sol record for historical billing and reproducibility. Use gpt-6.1-sol deliberately for new evaluations and migrations.

FAQ

What is the GPT-6.1 Sol model ID?

The direct OpenAI API model ID is gpt-6.1-sol.

How much does GPT-6.1 Sol cost?

Standard pricing is $2 per million input tokens, $0.10 per million cached input tokens, $2.50 per million cache-write tokens, and $10 per million output tokens. Long-context and alternative processing modes change those rates.

Is GPT-6.1 Sol the same as GPT-6 Sol?

No. They are separate versioned model IDs. GPT-6.1 Sol lowers cached-input pricing and has different documented reasoning and Chat Completions tool boundaries.

Does GPT-6.1 Sol support tools?

Yes through the Responses API, including functions, search, computer use, hosted shell, skills, MCP, and tool search. Chat Completions does not expose the same tool behavior for this model.

Is GPT-6.1 Sol available in GitHub Copilot?

GitHub announced a gradual rollout for Pro+, Max, Business, and Enterprise across its supported Copilot surfaces. Managed organizations may need an administrator to enable the model policy.