Model Details
GPT-5.1
Quick answer: GPT-5.1 is OpenAI's iterative update to GPT-5, continuing the August 2025 flagship with improvements to instruction following, agentic coding reliability, and long-context reasoning. It maintains the 400K-token context window and $1.25/$10 per million token pricing of GPT-5.
What GPT-5.1 Is
GPT-5.1 follows OpenAI's versioned checkpoint pattern — periodic updates to the GPT-5 family that improve specific capability areas without a full architectural change. The .1 update focuses on the agentic and tool-use capabilities that GPT-5 introduced, with refinements that reduce error rates in multi-step workflows and improve adherence to complex system prompts.
For Benchgen users, GPT-5.1 represents the same fundamental reasoning architecture as GPT-5, with the same router system (fast model + thinking model) and pricing, but with measurable improvements in the tasks most relevant to agent evaluation: coding precision, tool-call reliability, and multi-turn context retention.
Specifications
| Field | Value |
|---|---|
| Organization | OpenAI |
| Context window | 400,000 tokens |
| Max output | 128,000 tokens |
| License | Proprietary |
| Base series | GPT-5 (August 2025) |
| Modality | Multimodal (text and vision) |
Pricing
| Input (per 1M tokens) | Output (per 1M tokens) | |
|---|---|---|
| OpenAI | $1.25 | $10.00 |
Related Models & Benchmarks
- Compare: GPT-5 · GPT-5.2 · Claude Opus 4.6
- Benchmarks: SWE-bench Verified
GPT-5.1 vs Alternatives
| Model | BrowseComp | GPQA-Diamond | SimpleQA | License |
|---|---|---|---|---|
| GPT-5.1 | 90.0% | 88.1% | 45.6% | Proprietary |
| GPT-5.1 Thinking | 90.0% | 88.1% | — | Proprietary |
| GPT-5.1 Instant | 90.0% | 88.1% | — | Proprietary |
GPT-5.1 base model vs Thinking vs Instant: same BrowseComp + GPQA-Diamond across all three. Differentiated by HLE and speed.
Frequently Asked Questions
What is GPT-5.1?
OpenAI's January 2026 model scoring 90.0% BrowseComp, 88.1% GPQA-Diamond, 45.6% SimpleQA. Proprietary.Scores from Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.