GPT-6.1 Sol
Near-Astra intelligence for a fifth of the price
Context window
1,050,000
Input / 1M tokens
$2.00
Output / 1M tokens
$10.00
Provider
OpenAI
Cached input $0.10 and cache writes $2.50 per 1M. Prompts over 272K input tokens cost 2x input and cache rates and 1.5x output for the full request. Fast mode costs 2x standard and Ultrafast 6x. Max output 128K tokens (max input 922K); knowledge cutoff April 30, 2026. · Data verified 2026-11-08
GPT-6.1 Sol is an upgrade to GPT-6 Sol that nearly matches GPT-6 Astra's intelligence on agentic coding, computer use, and professional work at one-fifth of Astra's standard input and output token prices. The model delivers significant improvements over GPT-6 Sol in complex professional tasks including code writing, debugging, document understanding, and multistep business workflows.
Capability index
Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.
How to access it
Available in the OpenAI API (gpt-6.1-sol) on the Chat Completions, Responses and Batch endpoints, on Amazon Bedrock (openai.gpt-6.1-sol), and in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. Access starts with a basic API tier. Since October 21, 2026 also in private preview in Google Cloud Model Garden for select enterprise customers.
Strengths
- ✓Matches GPT-6 Astra on DeepSWE v1.1 coding tasks
- ✓Cost-efficient alternative to GPT-6 Astra at 1/5 the price
- ✓Improved factual accuracy with 32% reduction in error rate vs GPT-6 Sol
- ✓Strong performance on long-horizon computer use tasks (OSWorld 2.0)
- ✓Better alignment with user intent and safety constraints
Best for developers who...
When to choose it (and when not to)
Reach for GPT-6.1 Sol when...
- →When you need Astra-level intelligence but have a 1/5 budget constraint
- →For repeated and long-running workloads where cost per task matters
- →When coding performance and computer use are primary requirements
Look elsewhere if...
- ✕If you require the absolute highest safety standards (GPT-6 Astra has 2.4% vs 4.3% stress-test failure rate)
- ✕When maximum reasoning on scientific research tasks is critical (Astra scores 68.1% vs Sol's lower score)
How to use it
- ›Use prompt caching for repeated context to maximize the $0.10 cached input rate
- ›Structure complex tasks into distinct steps for better multi-step workflow handling
- ›Use higher reasoning effort (High/Max) for scientific or complex analysis work
Quickstart
Pythonfrom openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gpt-6.1-sol",
messages=[{"role": "user", "content": "Write a Python function to sort a list"}]
)
print(response.choices[0].message.content)API key required; available via OpenAI API with proper authentication.
API model id: gpt-6.1-sol
Benchmarks
| Benchmark | Score | Notes |
|---|---|---|
| DeepSWE v1.1 | 75.2% (high effort) | OpenAI-reported, as relayed by Vellum; 71.9% at max effort. GPT-6 Sol scored 68.8% |
| OSWorld 2.0 offline subset | 71.4% (max effort) | OpenAI-reported partial score, as relayed by Vellum; GPT-6 Sol scored 60.5% |
| Terminal-Bench Science 0.1 (Max Effort) | Cost: $5.47/task | Task-based pricing; Astra costs $23.80, Opus 5.5 costs $23.21 |
| AutomationBench (Medium Effort) | 2.2 points above Claude Opus 5.5 | Professional automation tasks |
| Factual Accuracy (Low Reasoning Effort) | 7.7% error rate | 32% reduction vs GPT-6 Sol's 11.4% error rate |
Source: OpenAI Official Announcement
Compare GPT-6.1 Sol
GPT-6.1 Sol vs Claude Fable 5.2
Anthropic - 1M ctx
GPT-6.1 Sol vs Claude Opus 5.5
Anthropic - 1M ctx
GPT-6.1 Sol vs Claude Sonnet 5.5
Anthropic - 1M ctx
GPT-6.1 Sol vs Gemini 4 Argon
Google - 1M (shared input + output) ctx
Compare GPT-6.1 Sol with any other model
Build a comparison →All model comparisons →