OpenAIGPT-4.1
Best configuration ranks #352 of 610 on the BenchLeader Index at 47.3 ±4.3. Last measured 2 Sept 2026. Released 14 Apr 2025.
- Blended price
- $3.50/M
- $2.00 in · $8.00 out
- Output speed
- 151 tok/s
- First answer
- 0.92 s
- first token 1.07 s
- Context
- 1.0M
- Overall index47
- Reasoning40
- Coding43
- Agents & tools58
- Maths41
- Knowledge40
- Instruction following35
- Human preference60
- Multimodal52
- Long context57
- Composite45
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| high | 46.2 | #370 | 151 tok/s | 0.92 s | $0.0032 | Coding 37 · Knowledge 55 · Maths 43 |
| defaultbest | 47.3 | #352 | 151 tok/s | 0.92 s | $0.0032 | Agents & tools 58 · Coding 43 · Composite 45 · Human preference 60 · Instruction following 35 · Knowledge 40 · Long context 57 · Maths 41 · Multimodal 52 · Reasoning 40 |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | high | default | Source |
|---|---|---|---|
| GPQA Diamond | – | 66.9%#163 | Epoch AI Benchmarking Hub |
| Humanity's Last Exam | – | 5.4%#38 | Scale AI / CAIS |
| SimpleBench | – | 27.0%#70 | SimpleBench |
| LMArena Hard Prompts | – | 1432#116 | LMArena |
| GPQA Diamond (AA)not in index | – | 66.6%#315 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | – | 4.2%#448 | Artificial Analysis |
| GPQA Diamond (Vals)not in index | 65.4%#98 | – | Vals AI |
| Kagi LLM Benchmark | – | 52.3%#73 | Kagi LLM Benchmark |
| ARC-AGI-1 | – | 5.5%#174 | ARC Prize |
| ARC-AGI-2 | – | 0.4%#171 | ARC Prize |
Coding
| Benchmark | high | default | Source |
|---|---|---|---|
| SWE-bench Verified (Epoch) | – | 48.5%#32 | Epoch AI Benchmarking Hub |
| WeirdML | – | 39.0%#107 | WeirdML |
| LMArena Coding | – | 1457#121 | LMArena |
| LiveCodeBench | 54.7%#109 | – | Vals AI |
| Aider Polyglot | – | 52.4%#23 | Aider polyglot leaderboard |
| SWE-bench Verified (bash only) | – | 39.6%#35 | SWE-bench |
| SWE-bench Verified (any scaffold)not in index | – | 39.6%#52 | SWE-bench |
Agents & tools
| Benchmark | high | default | Source |
|---|---|---|---|
| Terminal-Bench Hard | – | 13.6%#194 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | – | 47.1%#197 | Artificial Analysis |
| BFCL Overall | – | 54.0%#18 | Berkeley Function Calling Leaderboard |
Maths
| Benchmark | high | default | Source |
|---|---|---|---|
| FrontierMath Tiers 1–3 | – | 6.0%#94 | Epoch AI Benchmarking Hub |
| OTIS Mock AIME | – | 38.3%#181 | Epoch AI Benchmarking Hub |
| MATH Level 5 | – | 83.0%#30 | Epoch AI Benchmarking Hub |
| AIME (Vals) | 39.6%#68 | – | Vals AI |
| MGSM | 87.7%#56 | – | Vals AI |
Knowledge
| Benchmark | high | default | Source |
|---|---|---|---|
| SimpleQA Verified | – | 31.1%#55 | Epoch AI Benchmarking Hub |
| AA-Omniscience | – | -39.6#253 | Artificial Analysis |
| MMLU-Pro | 80.5%#84 | – | Vals AI |
| LegalBench | 83.1%#53 | – | Vals AI |
| CorpFin | 63.0%#46 | – | Vals AI |
| TaxEval | 75.1%#23 | – | Vals AI |
| MedQA | 91.2%#39 | – | Vals AI |
| PRBench Finance | – | 34.3%#30 | Scale AI SEAL |
| PRBench Legal | – | 36.5%#29 | Scale AI SEAL |
| MultiNRC | – | 21.2%#32 | Scale AI SEAL |
Instruction following
| Benchmark | high | default | Source |
|---|---|---|---|
| IFBench | – | 43.0%#225 | Artificial Analysis |
| MultiChallenge | – | 39.4%#27 | Scale AI SEAL |
Human preference
| Benchmark | high | default | Source |
|---|---|---|---|
| LMArena Text | – | 1414#113 | LMArena |
Multimodal
| Benchmark | high | default | Source |
|---|---|---|---|
| LMArena Vision | – | 1210#68 | LMArena |
| MMMU-Pro | – | 61.2%#173 | Artificial Analysis |
| VISTA | – | 45.3%#26 | Scale AI SEAL |
Long context
| Benchmark | high | default | Source |
|---|---|---|---|
| Fiction.LiveBench 120k | – | 62.5%#13 | Fiction.live |
| AA-LCR | – | 68.3%#178 | Artificial Analysis |
Composite
| Benchmark | high | default | Source |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | 136.8#111 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | – | 12.7#259 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| OpenAI | 54 tok/s | 1.07 s | $2.00 | $8.00 | 1.0M | – |
| Azure | 20 tok/s | 1.82 s | $2.00 | $8.00 | 1.0M | – |
| Azure (EU) | 20 tok/s | 0.73 s | $2.20 | $8.80 | 1.0M | – |
Price history
Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.500 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0032 | $0.0028 | 2.9 s |
| Summarise a 30-page report | 12,000 / 600 | $0.029 | $0.015 | 4.9 s |
| Code edit | 6,000 / 1,500 | $0.024 | $0.017 | 10.9 s |
| Agentic coding session | 60,000 / 4,000 | $0.152 | $0.085 | 27.5 s |
| Structured extraction | 2,000 / 200 | $0.0056 | $0.0034 | 2.2 s |
See also
Data as of 9 Sept 2026. Compare these configurations.