What is GPT-6.1 Sol?
GPT-6.1 Sol is OpenAI’s upgraded mid-tier GPT-6 model for coding, computer use and professional work. Released September 29, 2026, it keeps the $2 input and $10 output token prices of GPT-6 Sol while improving capability and cutting cached-input pricing in half.
OpenAI positions the model between GPT-6 Astra and GPT-6 Luna. Astra remains the flagship for the hardest work, while Luna targets inexpensive high-volume tasks. GPT-6.1 Sol is intended to capture much of Astra’s practical capability at a lower cost and with broader room for iteration.
QUICK VERDICT: GPT-6.1 Sol is the strongest cost-to-capability option in OpenAI’s current GPT-6 family for many complex workloads. It approaches Astra on several company evaluations at one-fifth of Astra’s standard token price, but the best results still depend on reasoning effort, tools and a carefully designed agent harness.
Field | Verified value |
|---|---|
Provider | OpenAI |
Release date | September 29, 2026 |
License | Proprietary |
API model ID | gpt-6.1-sol |
Context window | 1,050,000 tokens |
Maximum input | 922,000 tokens |
Maximum output | 128,000 tokens |
Modalities | Text and image input; text output |
Knowledge cutoff | April 30, 2026 |
Reasoning effort | Low, medium, high, xhigh and max; medium is default |
Base API pricing | $2 input, $0.10 cached input and $10 output per million tokens |
Best fit | Complex coding, computer use and professional workflows |
Last verified | October 2, 2026 |
What changed from GPT-6 Sol
GPT-6.1 Sol is a capability and efficiency update rather than a new price tier. It raises performance across coding, document work, business automation, computer use and scientific tasks while preserving the base Sol input and output rates. The cached-input rate drops from $0.20 to $0.10 per million tokens.
Area | GPT-6.1 Sol change |
|---|---|
Coding | OpenAI reports a 6.4-point gain over GPT-6 Sol’s best DeepSWE result |
Computer use | Seven-point gain over GPT-6 Sol on the cited OSWorld setting |
Professional work | Higher scores on document and multi-tool business workflows |
Scientific work | More than double GPT-6 Sol’s cited Terminal-Bench Science result at max effort |
Cached input | $0.10 per million tokens instead of $0.20 |
Reasoning | No none or minimal setting; medium is the default |
The practical question is not whether GPT-6.1 Sol is newer. It is whether the change improves accepted-task quality, speed and cost in the exact workflow. Keep prompts, tools, source material and scoring criteria fixed when comparing it with GPT-6 Sol.
Capabilities and best use cases
The model supports web search, file search, image generation, code execution, hosted shell, computer use, MCP connections and other tools through the Responses API. Chat Completions is supported without tool calling, so agent builders should use Responses when they need the full tool set.
Capability | Practical use |
|---|---|
Coding | Repository work, debugging, implementation and software-engineering agents |
Computer use | GUI workflows through the supported computer-use tool |
Professional documents | PDF, chart, table and fine-print analysis |
Business automation | Multi-step work across sales, marketing, operations, support, finance and HR |
Scientific workflows | Code, terminal, data analysis, simulation and theorem-oriented tasks |
Long context | Large repositories and document collections up to the documented input limit |
The strongest documented fit is complex coding, computer use and professional work when Astra is unnecessarily expensive. A capable base model can still fail when retrieval is weak, tools are over-permissioned, instructions conflict or the system lacks verification. Evaluate the full application rather than the model in isolation.
Benchmark evidence
OpenAI reports that GPT-6.1 Sol matches GPT-6 Astra on its cited DeepSWE setting at roughly one-fifth of Astra’s cost. It also reports gains over GPT-6 Sol on AutomationBench, OSWorld and Terminal-Bench Science. These are provider evaluations and should be validated with the same tools, effort settings and acceptance criteria used in production.
Evaluation | Reported result | Evidence boundary |
|---|---|---|
DeepSWE v1.1 | Matches Astra in the cited comparison | 6.4 points above GPT-6 Sol’s best cited result |
AutomationBench 1.0.6 | 2.2 points above Opus 5.5 at medium effort | Provider comparison across 47 business tools |
OSWorld 2.0 offline | Seven points above GPT-6 Sol at max effort | Within 2.1 points of Astra in the cited setting |
Terminal-Bench Science 0.1 | More than double GPT-6 Sol at max effort | OpenAI reports $5.47 average cost per task |
Factuality evaluation | 7.7% error-containing responses at low effort | Down from 11.4% for GPT-6 Sol on difficult flagged conversations |
Benchmark scores depend on model version, reasoning effort, sampling, tools, scaffolding and the exact dataset release. Do not combine scores from different harnesses into a synthetic ranking. Use public results to choose candidates, then test those candidates on frozen tasks from the intended workflow.
Artificial Analysis Intelligence Index
Artificial Analysis independently measures model intelligence, speed and cost. GPT-6.1 Sol scores 52 on the Artificial Analysis Intelligence Index at max effort. 1 point below GPT-6 Astra at less than one quarter of Astra’s cost per task.
Metric | Value |
|---|---|
Intelligence Index score | 52 |
Effort setting | max |
Source | Artificial Analysis (independent) |
Last verified | October 2, 2026 |
The Artificial Analysis Intelligence Index combines scores across mathematics, reasoning, coding, instruction following and language tasks. The scale is not a percentage: the index reflects relative position across the models they track, not a share of correct answers. Compare scores only within the same effort setting and the same index version.
Pricing and access
GPT-6.1 Sol is available through the OpenAI API, ChatGPT Work and Codex for eligible paid users. At launch it was not yet available in ordinary Chat. The model supports US and EU data residency, although Fast mode is unavailable with EU data residency.
Long prompts above 272,000 input tokens trigger higher rates for the entire request: double input and cache pricing and 1.5 times output pricing. Batch and Flex processing are 50% below Standard rates, while Fast mode is twice Standard pricing.
Usage or access item | Current value |
|---|---|
Standard input | $2 per million tokens |
Cached input | $0.10 per million tokens |
Cache writes | $2.50 per million tokens |
Standard output | $10 per million tokens |
Batch and Flex | 50% below Standard token rates |
Fast mode | 2 times Standard rates |
Prompts above 272K | 2 times input and cache; 1.5 times output for the full request |
Token price is only one part of total cost. Include cache behavior, long-context multipliers, tool calls, retries, wall time, failed-task recovery and human review. The useful comparison is cost per accepted result, not price per million tokens in isolation.
Limitations and deployment risks
- The lowest reasoning settings, none and minimal, are not supported.
- Prompts above 272,000 tokens incur long-context multipliers across the full request.
- The strongest provider results often use high or maximum reasoning effort and tool-enabled harnesses.
- Chat Completions does not provide the complete tool-calling path available in the Responses API.
- A 1.05-million-token window does not guarantee accurate retrieval or instruction retention across every token.
- Computer-use and shell tools require least-privilege permissions, approval gates and recoverable actions.
- OpenAI’s benchmark comparisons are useful but are not independent production tests.
High-stakes legal, medical, financial, scientific and security work needs qualified review. Log the exact model version and configuration, separate untrusted content from system instructions, restrict tools to the minimum required scope and require approval before irreversible or externally visible actions.
How to evaluate GPT-6.1 Sol
Build a frozen evaluation set of 20 to 50 real tasks. Include routine work, difficult edge cases, adversarial inputs, long-context examples and cases where the correct behavior is to stop or escalate. Compare GPT-6.1 Sol with GPT-6 Sol and at least one neighboring model using equivalent tools and source material.
Test area | What to record |
|---|---|
Task completion | Pass or fail against a written acceptance rubric |
Reliability | Repeated-run success rate, variance and silent failures |
Factuality | Unsupported claims, source use, quotations and citation accuracy |
Tool use | Wrong calls, retries, recovery and permission-boundary failures |
Long context | Retrieval accuracy, instruction retention and cost at realistic lengths |
Efficiency | Wall time, input, output, cached tokens, tool charges and review time |
Safety | Prompt injection, sensitive-data handling and irreversible-action controls |
Choose the least expensive configuration that clears the acceptance threshold with a safety margin. Re-run the suite when the model, system prompt, effort level, retrieval layer, tool definitions or approval policy changes.
Who should use it?
Situation | Recommendation |
|---|---|
Strong fit | complex coding, computer use and professional work when Astra is unnecessarily expensive |
Pilot first | Long-running agents, large contexts, computer use and workflows with several tools |
Escalate | Ambiguous or consequential work that does not reliably clear the evaluation threshold |
Avoid unsupervised use | Irreversible actions, sensitive data or high-stakes decisions without monitoring and approval |
A migration should be driven by measured outcomes. Keep GPT-6 Sol available during the pilot, record where each model succeeds or fails, and use routing when different task classes have different quality and cost requirements.
Frequently asked questions
Is GPT-6.1 Sol open source?
No. GPT-6.1 Sol is proprietary. Access, serving behavior and lifecycle decisions are controlled by OpenAI and supported distribution partners.
How much does GPT-6.1 Sol cost?
$2 per million tokens. Review the full pricing table above because cached input, long context, processing mode and platform can materially change the total.
What is GPT-6.1 Sol best used for?
Its strongest documented fit is complex coding, computer use and professional work when Astra is unnecessarily expensive. Start with a supervised pilot and retain human sign-off for consequential work.
Should I migrate from GPT-6 Sol?
Only after a side-by-side evaluation. Measure accepted-task quality, total cost, latency, output style, tool reliability and migration engineering. A newer model is not automatically the better operational choice.
How should benchmark evidence for GPT-6.1 Sol be interpreted?
Use provider results to identify promising workloads, then reproduce the comparison with the exact model configuration, tools and acceptance criteria that matter to the deployment. Do not infer a site ranking from benchmark claims gathered under a different harness.
Related model guides
Browse the Best AI Models directory.
Compare with the GPT-6 Astra guide.
Compare with the GPT 5.6 Sol guide.
Compare with the Claude Opus 5.5 guide.
Official sources and update policy
OpenAI’s GPT-6.1 Sol announcement.
GPT-6.1 Sol API model documentation.
Checked October 2, 2026. We update this guide when OpenAI changes the specification, pricing, access, safety documentation or model lifecycle. Vendor benchmarks are attributed and are not presented as independent testing. No vendor payment or affiliate relationship determined inclusion.