OpenAIGPT-4o mini
GPT-4o mini is an OpenAI proprietary model, released 18 Jul 2024. Its best configuration ranks #758 of 785 on the BenchLeader Index at 35.8 ±4.0, in the lower half. The ± is the point: 138 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in human preference (47) and lowest in knowledge (23). At $0.262 per million tokens blended it is cheaper than most ranked models. Output speed of 103 tokens per second puts it faster than most, with a first answer in 1.1 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Oct 2026.
- Blended price
- $0.262/M
- $0.150 in · $0.600 out
- Output speed
- 103 tok/s
- measured by Artificial Analysis
- First answer
- 1.13 s
- first token 0.56 s
- Context
- 128k
- Full answer
- 5.97 s
- median, reasoning included
- Cost per run
- –
- Released
- 18 Jul 2024
- Overall index36
- Reasoning30
- Coding29
- Agents & tools30
- Maths37
- Knowledge23
- Instruction following42
- Human preference47
- Multimodal32
- Composite37
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. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| high | 33.7 | #776 | 103 tok/s | 1.13 s | $0.0002 | Coding 16 · Knowledge 29 · Maths 35 |
| not statedbest | 35.8 | #758 | 103 tok/s | 1.13 s | $0.0002 | Agents & tools 30 · Coding 29 · Composite 37 · Human preference 47 · Instruction following 42 · Knowledge 23 · Maths 37 · Multimodal 32 · Reasoning 30 |
How its index has moved
8 Sept 2026 to 21 Oct 2026- not stated
- high
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. 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 | not stated | Source | Trend |
|---|---|---|---|---|
| GPQA Diamondeffort not stated | – | 37.7%#254 | Epoch AI Benchmarking Hub | |
| SimpleBencheffort not stated | – | 10.7%#87 | SimpleBench | |
| Arena Hard Promptseffort not stated | – | 1311#252 | Arena (LMArena) | |
| GPQA Diamond (AA)not in indexeffort not stated | – | 42.6%#466 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in indexeffort not stated | – | 4.2%#526 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | 44.2%#122 | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | 28.8%#125 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | 36.8%#53 | HELM Capabilities | |
| ARC-AGI-2effort not stated | – | 0.0%#242 | ARC Prize | |
| Chess Puzzlesnot in indexeffort not stated | – | 3.0%#135 | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in indexeffort not stated | – | 12.0%#103 | Epoch AI Benchmarking Hub | |
| LMCAnot in indexeffort not stated | – | 10.4%#177 | Epoch AI Benchmarking Hub | |
| DTBenchnot in indexeffort not stated | – | 54.4%#198 | Epoch AI Benchmarking Hub | |
| PIQAnot in indexeffort not stated | – | 88.7%#1 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| WeirdMLeffort not stated | – | 11.8%#148 | WeirdML | |
| Arena Codingeffort not stated | – | 1348#245 | Arena (LMArena) | |
| LiveCodeBench | 26.4%#133 | – | Vals AI | |
| Aider Polygloteffort not stated | – | 3.6%#48 | Aider polyglot leaderboard | |
| Terminal-Bench 2.1 (AA)not in indexeffort not stated | – | 5.6%#203 | Artificial Analysis |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| τ³-Banking (AA) | 2.9%#199 | Artificial Analysis | |
| BALROGnot in index | 17.4%#27 | BALROG | |
| BrowseCompnot in index | 0.6%#23 | OpenAI |
Maths
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| FrontierMath Tiers 1–3effort not stated | – | 0.7%#115 | Epoch AI Benchmarking Hub | |
| OTIS Mock AIMEeffort not stated | – | 6.9%#238 | Epoch AI Benchmarking Hub | |
| MATH Level 5effort not stated | – | 52.6%#52 | Epoch AI Benchmarking Hub | |
| AIME (Vals) | 11.5%#84 | – | Vals AI | |
| MGSM | 86.2%#60 | – | Vals AI | |
| Omni-MATH (HELM)effort not stated | – | 28.0%#48 | HELM Capabilities | |
| MATH 500not in index | 72.6%#46 | – | Vals AI | |
| Arena Mathseffort not stated | – | 1275#250 | Arena (LMArena) | |
| GSM8Knot in indexeffort not stated | – | 91.3%#5 | Epoch AI Benchmarking Hub |
Knowledge
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| SimpleQA Verifiedeffort not stated | – | 8.3%#86 | Epoch AI Benchmarking Hub | |
| MMLU-Pro | 62.7%#126 | – | Vals AI | |
| CorpFin | 45.5%#109 | – | Vals AI | |
| TaxEval | 60.5%#122 | – | Vals AI | |
| MedQA | 72.4%#77 | – | Vals AI | |
| MMLU-Pro (HELM)not in indexeffort not stated | – | 60.3%#47 | HELM Capabilities | |
| MMLUnot in indexeffort not stated | – | 81.8%#16 | Epoch AI Benchmarking Hub |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 30.9%#360 | Artificial Analysis | |
| IFEval (HELM) | 78.2%#52 | HELM Capabilities | |
| WildBench (HELM) | 79.1%#39 | HELM Capabilities | |
| Arena Instruction Followingnot in index | 1293#244 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1318#238 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1294#221 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1311#236 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1326#229 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 67.2%#35 | Lech Mazur |
Multimodal
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Visioneffort not stated | – | 1066#131 | Arena (LMArena) | |
| MMMU-Proeffort not stated | – | 41.5%#264 | Artificial Analysis | |
| MMMU (validation)effort not stated | – | 59.4%#34 | MMMU | |
| MMMU-Pro (official)not in indexeffort not stated | – | 37.6%#24 | MMMU | |
| MMMU-Pro (Vals)not in index | 56.6%#85 | – | Vals AI | |
| MortgageTaxnot in index | 54.5%#75 | – | Vals AI | |
| GeoBenchnot in indexeffort not stated | – | 64.0%#14 | GeoBench | |
| VPCTnot in indexeffort not stated | – | 34.0%#27 | Epoch AI Benchmarking Hub | |
| Video-MMEnot in indexeffort not stated | – | 64.8%#17 | Video-MME |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | 126.5#159 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 6.7#516 | Artificial Analysis | |
| HELM Capabilities meannot in index | 56.5%#51 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 48.1%#14 | Scale AI SEAL |
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 | 78 tok/s | 0.51 s | $0.150 | $0.600 | 128k | – |
| Azure | 76 tok/s | 0.68 s | $0.165 | $0.660 | 128k | – |
| Azure | 72 tok/s | 1.11 s | $0.150 | $0.600 | 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 $0.075 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0002 | $0.0002 | 4.0 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0022 | $0.0015 | 6.9 s |
| Code edit | 6,000 / 1,500 | $0.0018 | $0.0015 | 15.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.011 | $0.0080 | 39.8 s |
| Structured extraction | 2,000 / 200 | $0.0004 | $0.0003 | 3.1 s |
See also
Questions people ask
- What are GPT-4o mini's benchmark scores?
- As of 25 Oct 2026, GPT-4o mini scores 35.8 ±4.0 on the BenchLeader Index, strongest in human preference (47) and weakest in knowledge (23). That rests on 23 independent benchmark results, and every score on this page links to its publisher.
- How does GPT-4o mini rank?
- GPT-4o mini is #758 of 785 ranked configurations on 25 Oct 2026. 138 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does GPT-4o mini cost per 1M tokens?
- As of 25 Oct 2026, GPT-4o mini costs $0.150 per 1M input tokens and $0.600 per 1M output tokens, $0.262 blended (three input tokens per output token).
- When was GPT-4o mini released?
- OpenAI released GPT-4o mini on 18 Jul 2024. BenchLeader has results for it at 2 reasoning-effort settings.
Data as of 25 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GPT-4o mini: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/gpt-4o-mini.