OpenAIGPT-4o
Best configuration ranks #461 of 610 on the BenchLeader Index at 42.3 ±5.4. Last measured 2 Sept 2026. Released 13 May 2024.
- Blended price
- $4.38/M
- $2.50 in · $10.00 out
- Output speed
- 140 tok/s
- First answer
- 1.04 s
- first token 0.77 s
- Context
- 128k
- Overall index42
- Reasoning37
- Coding35
- Agents & tools31
- Maths32
- Knowledge45
- Instruction following29
- Human preference64
- Multimodal49
- Long context54
- Composite40
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 | 41.1 | #493 | 100 tok/s | 1.20 s | $0.0065 | Coding 28 · Knowledge 46 · Maths 40 |
| defaultbest | 42.3 | #461 | 140 tok/s | 1.04 s | $0.0040 | Agents & tools 31 · Coding 35 · Composite 40 · Human preference 64 · Instruction following 29 · Knowledge 45 · Long context 54 · Maths 32 · Multimodal 49 · Reasoning 37 |
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 | – | 49.2%#201 | Epoch AI Benchmarking Hub |
| Humanity's Last Exam | – | 2.7%#44 | Scale AI / CAIS |
| SimpleBench | – | 17.8%#83 | SimpleBench |
| LMArena Hard Prompts | – | 1456#88 | LMArena |
| GPQA Diamond (AA)not in index | – | 65.5%#327 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | – | 4.0%#470 | Artificial Analysis |
| GPQA Diamond (Vals)not in index | 53.8%#113 | – | Vals AI |
| Kagi LLM Benchmark | – | 75.0%#12 | Kagi LLM Benchmark |
| ARC-AGI-1 | – | 4.5%#179 | ARC Prize |
| ARC-AGI-2 | – | 0.0%#174 | ARC Prize |
Coding
| Benchmark | high | default | Source |
|---|---|---|---|
| SWE-bench Verified (Epoch) | – | 31.0%#33 | Epoch AI Benchmarking Hub |
| WeirdML | – | 25.1%#130 | WeirdML |
| GSO-Bench | – | 0.0%#31 | GSO-Bench |
| LMArena Coding | – | 1468#103 | LMArena |
| LiveCodeBench | 43.4%#118 | – | Vals AI |
| Aider Polyglot | – | 45.3%#26 | Aider polyglot leaderboard |
| SWE-bench Verified (bash only) | – | 21.6%#40 | SWE-bench |
| SWE-bench Verified (any scaffold)not in index | – | 38.8%#53 | SWE-bench |
Agents & tools
| Benchmark | high | default | Source |
|---|---|---|---|
| GDPval | – | 9.9%#11 | OpenAI |
| Cybench | – | 12.5%#14 | Cybench |
| APEX-Agents | – | 1.1%#62 | Mercor |
| Terminal-Bench Hard | – | 8.3%#226 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | – | 28.9%#258 | Artificial Analysis |
Maths
| Benchmark | high | default | Source |
|---|---|---|---|
| FrontierMath Tiers 1–3 | – | 0.3%#102 | Epoch AI Benchmarking Hub |
| OTIS Mock AIME | – | 6.4%#222 | Epoch AI Benchmarking Hub |
| MATH Level 5 | – | 53.3%#50 | Epoch AI Benchmarking Hub |
| AIME (Vals) | 14.0%#83 | – | Vals AI |
| MGSM | 90.7%#36 | – | Vals AI |
Knowledge
| Benchmark | high | default | Source |
|---|---|---|---|
| SimpleQA Verified | – | 26.0%#59 | Epoch AI Benchmarking Hub |
| AA-Omniscience | – | -10.5#142 | Artificial Analysis |
| MMLU-Pro | 74.1%#109 | – | Vals AI |
| LegalBench | 82.2%#66 | – | Vals AI |
| CorpFin | 45.9%#109 | – | Vals AI |
| TaxEval | 74.5%#34 | – | Vals AI |
| MedQA | 88.2%#54 | – | Vals AI |
| MultiNRC | – | 12.4%#39 | Scale AI SEAL |
Instruction following
| Benchmark | high | default | Source |
|---|---|---|---|
| IFBench | – | 36.0%#293 | Artificial Analysis |
| TutorBench | – | 36.1%#27 | Scale AI SEAL |
Human preference
| Benchmark | high | default | Source |
|---|---|---|---|
| LMArena Text | – | 1443#75 | LMArena |
Multimodal
| Benchmark | high | default | Source |
|---|---|---|---|
| LMArena Vision | – | 1244#55 | LMArena |
| MMMU-Pro | – | 56.3%#190 | Artificial Analysis |
| VISTA | – | 38.0%#42 | Scale AI SEAL |
| MMMU (validation) | – | 70.7%#21 | MMMU |
| MMMU-Pro (official)not in index | – | 51.9%#15 | MMMU |
Long context
| Benchmark | high | default | Source |
|---|---|---|---|
| AA-LCR | – | 56.0%#231 | Artificial Analysis |
Composite
| Benchmark | high | default | Source |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | 129.0#137 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | – | 9.0#343 | 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 |
|---|---|---|---|---|---|---|
| Azure | 63 tok/s | 0.88 s | $2.50 | $10.00 | 128k | – |
| OpenAI | 48 tok/s | 0.67 s | $2.50 | $10.00 | 128k | – |
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 $1.25 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0040 | $0.0036 | 3.2 s |
| Summarise a 30-page report | 12,000 / 600 | $0.036 | $0.025 | 5.3 s |
| Code edit | 6,000 / 1,500 | $0.030 | $0.024 | 11.8 s |
| Agentic coding session | 60,000 / 4,000 | $0.190 | $0.134 | 29.7 s |
| Structured extraction | 2,000 / 200 | $0.0070 | $0.0051 | 2.5 s |
See also
Data as of 9 Sept 2026. Compare these configurations.