OpenAIGPT-5 mini
GPT-5 mini is an OpenAI proprietary reasoning model, released 7 Aug 2025. Its best configuration ranks #202 of 785 on the BenchLeader Index at 56.5 ±7.6, in the upper half. The ± is the point: 360 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 maths (79) and lowest in knowledge (40). At $0.688 per million tokens blended it is mid-priced. Output speed of 86 tokens per second puts it slower than most, with a first token in 3.2 s. It has been measured at 6 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 17 Feb 2026.stale: no new result in six months
- Blended price
- $0.688/M
- $0.250 in · $2.00 out
- Output speed
- 86 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 3.16 s
- Context
- 400k
- Full answer
- –
- Cost per run
- –
- Released
- 7 Aug 2025
- Overall index57
- Reasoning54
- Coding47
- Agents & tools55
- Maths79
- Knowledge40
- Instruction following69
- Multimodal64
Versions
OpenAI has shipped 3 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| GPT-5.4 Minixhigh | 17 Mar 2026 | 56.3 | #206 |
| GPT-5 minithis page | 7 Aug 2025 | 56.5 | #202 |
| GPT-4.1 minihigh | 14 Apr 2025 | 46.3 | #492 |
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 |
|---|---|---|---|---|---|---|
| minimal | 43.1 | #578 | 121 tok/s | 0.81 s | $0.0007 | Agents & tools 46 · Composite 41 · Instruction following 48 · Knowledge 38 · Long context 44 · Maths 35 · Multimodal 41 · Reasoning 39 |
| low | – | – | 86 tok/s | 3.16 s | $0.0007 | Maths 35 · Reasoning 33 |
| medium | 53.8 | #269 | 121 tok/s | 9.45 s | $0.0007 | Agents & tools 47 · Coding 49 · Composite 53 · Instruction following 70 · Knowledge 57 · Long context 58 · Maths 64 · Multimodal 52 · Reasoning 43 |
| high | 54.1 | #260 | 119 tok/s | 52 s | $0.0007 | Agents & tools 50 · Coding 54 · Composite 48 · Human preference 56 · Instruction following 74 · Knowledge 51 · Long context 61 · Maths 53 · Multimodal 52 · Reasoning 45 |
| thinking | – | – | 86 tok/s | 3.16 s | $0.0007 | Instruction following 50 |
| not statedbest | 56.5 | #202 | 86 tok/s | 3.16 s | $0.0007 | Agents & tools 55 · Coding 47 · Instruction following 69 · Knowledge 40 · Maths 79 · Multimodal 64 · Reasoning 54 |
How its index has moved
8 Sept 2026 to 21 Oct 2026- not stated
- high
- medium
- minimal
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 | minimal | low | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| GPQA Diamond | 71.7%#163 | – | 71.7%#166 | 75.0%#151 | – | Epoch AI Benchmarking Hub | |
| Humanity's Last Exameffort not stated | – | – | – | – | 19.4%#22 | Scale AI / CAIS | |
| Arena Hard Prompts | – | – | – | 1404#169 | – | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 68.7%#310 | – | 80.3%#192 | 82.8%#166 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 5.0%#451 | – | 15.9%#245 | 21.5%#208 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | – | 80.3%#65 | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | – | – | 70.3%#26 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | – | – | – | 75.6%#3 | HELM Capabilities | |
| ARC-AGI-1 | 5.3%#246 | 26.3%#212 | 37.3%#197 | 54.3%#174 | – | ARC Prize | |
| ARC-AGI-2 | 1.7%#210 | 0.8%#231 | 4.0%#185 | 4.4%#183 | – | ARC Prize | |
| CritPt | 0.0%#315 | – | 1.4%#211 | 0.0%#315 | – | Artificial Analysis | |
| Chess Puzzlesnot in index | 0.0%#145 | 16.0%#104 | – | 24.0%#80 | – | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in index | 10.0%#112 | – | 4.0%#137 | 5.0%#132 | – | Epoch AI Benchmarking Hub | |
| LMCAnot in index | – | – | – | 34.2%#128 | – | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | – | – | 80.5%#128 | – | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | – | – | 63.3%#44 | ForecastBench |
Coding
| Benchmark | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|
| SWE-bench Verified (Epoch) | 64.7%#27 | – | – | Epoch AI Benchmarking Hub | |
| SciCode | – | 39.0%#169 | 39.2%#167 | SciCode | |
| WeirdML | – | 52.7%#63 | – | WeirdML | |
| Arena Coding | – | 1432#172 | – | Arena (LMArena) | |
| SciCode (AA)not in index | – | 39.0%#181 | – | Artificial Analysis | |
| LiveCodeBench | – | 86.6%#20 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | 60.8%#74 | – | Vals AI | |
| SWE-bench Verified (bash only) | 59.8%#24 | – | 56.2%#27 | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | 59.8%#31 | – | – | SWE-bench | |
| Vibe Code Bench v1.1not in index | – | 14.2%#91 | – | Vals AI | |
| ALE-Benchnot in index | – | 799.8#73 | – | Epoch AI Benchmarking Hub | |
| AlgoTunenot in index | – | 1.4#15 | – | AlgoTune | |
| Terminal-Bench 4.0 (AA)not in index | – | 0.0%#152 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | – | 3.8%#213 | – | Artificial Analysis |
Agents & tools
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Terminal-Bench | – | 31.9%#59 | – | 34.8%#56 | Terminal-Bench | |
| Terminal-Bench Hard | 14.4%#196 | 28.8%#117 | 33.3%#92 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 31.9%#249 | 71.0%#153 | 68.4%#160 | – | Artificial Analysis | |
| BFCL Overalleffort not stated | – | – | – | 55.5%#15 | Berkeley Function Calling Leaderboard | |
| τ³-Banking (AA) | – | – | 15.5%#116 | – | Artificial Analysis | |
| Terminal-Bench 2.0 (Vals)not in index | – | – | 27.0%#49 | – | Vals AI | |
| Vending-Bench 2not in indexeffort not stated | – | – | – | -31.2#69 | Andon Labs | |
| AutomationBenchnot in index | – | – | 6.5%#160 | – | Artificial Analysis | |
| GDP.pdf (AA)not in index | – | – | 8.4%#142 | – | Artificial Analysis |
Maths
| Benchmark | minimal | low | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 6.0%#109 | 18.3%#95 | – | 46.7%#55 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | 12.2%#62 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 55.6%#174 | – | 78.3%#121 | 86.7%#86 | – | Epoch AI Benchmarking Hub | |
| MATH Level 5 | – | – | 96.8%#8 | 97.8%#3 | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | 9.0%#66 | – | Vals AI | |
| AIME (Vals) | – | – | – | 91.5%#24 | – | Vals AI | |
| MGSM | – | – | – | 92.6%#15 | – | Vals AI | |
| MathArena Apex | – | – | – | 1.0%#35 | – | MathArena | |
| Omni-MATH (HELM)effort not stated | – | – | – | – | 72.2%#1 | HELM Capabilities | |
| MATH 500not in index | – | – | – | 94.8%#6 | – | Vals AI | |
| Arena Maths | – | – | – | 1407#141 | – | Arena (LMArena) |
Knowledge
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | 21.6%#72 | – | Epoch AI Benchmarking Hub | |
| AA-Omniscience | -53.0#415 | -11.0#205 | -17.3#232 | – | Artificial Analysis | |
| MMLU-Pro | – | – | 82.2%#74 | – | Vals AI | |
| LegalBench | – | – | 81.8%#79 | – | Vals AI | |
| CorpFin | – | – | 60.2%#70 | – | Vals AI | |
| TaxEval | – | – | 75.2%#21 | – | Vals AI | |
| MedQA | – | – | 96.1%#6 | – | Vals AI | |
| MultiNRCeffort not statedpossible contamination | – | – | – | 23.9%#31 | Scale AI SEAL | |
| MMLU-Pro (HELM)not in indexeffort not stated | – | – | – | 83.5%#12 | HELM Capabilities | |
| CaseLaw v2not in index | – | – | 68.5%#4 | – | Vals AI | |
| MedCodenot in index | – | – | 43.0%#47 | – | Vals AI | |
| MedScribenot in index | – | – | 80.6%#54 | – | Vals AI | |
| AA-Omniscience: accuracynot in index | 19.2%#315 | 23.1%#264 | 25.0%#233 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 10.7%#406 | 55.7%#94 | 43.6%#147 | – | Artificial Analysis |
Instruction following
| Benchmark | minimal | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| IFBench | 45.6%#202 | 71.2%#60 | 75.4%#31 | – | – | Artificial Analysis | |
| MultiChallenge | – | – | – | 59.0%#15 | – | Scale AI SEAL | |
| IFEval (HELM)effort not stated | – | – | – | – | 92.7%#6 | HELM Capabilities | |
| WildBench (HELM)effort not stated | – | – | – | – | 85.5%#8 | HELM Capabilities | |
| Arena Instruction Followingnot in index | – | – | 1373#171 | – | – | Arena (LMArena) |
Human preference
| Benchmark | medium | high | Source | Trend |
|---|---|---|---|---|
| Arena Text | – | 1390#163 | Arena (LMArena) | |
| Arena Creative Writingnot in index | – | 1323#192 | Arena (LMArena) | |
| Arena Multi-turnnot in index | – | 1373#178 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | – | 1370#185 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 83.1%#8 | – | Lech Mazur |
Multimodal
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Arena Vision | – | – | 1202#90 | – | Arena (LMArena) | |
| MMMU-Pro | 58.4%#218 | 68.8%#160 | 70.1%#148 | – | Artificial Analysis | |
| VISTAeffort not stated | – | – | – | 50.4%#9 | Scale AI SEAL | |
| MMMU-Pro (Vals)not in index | – | – | 78.9%#49 | – | Vals AI | |
| MortgageTaxnot in index | – | – | 66.9%#30 | – | Vals AI | |
| SAGEnot in index | – | – | 43.0%#52 | – | Vals AI | |
| VPCTnot in index | – | 40.2%#15 | 39.0%#19 | – | Epoch AI Benchmarking Hub |
Long context
| Benchmark | minimal | medium | high | Source | Trend |
|---|---|---|---|---|---|
| Fiction.LiveBench 120k | – | 62.5%#13 | – | Fiction.live | |
| AA-LCR | 39.3%#362 | 72.3%#188 | 72.3%#188 | Artificial Analysis | |
| MLCRnot in index | – | – | 8.9%#122 | Artificial Analysis |
Composite
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | – | 145.3#84 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 9.9#386 | 20.6#227 | 16.8#268 | – | Artificial Analysis | |
| HELM Capabilities meannot in indexeffort not stated | – | – | – | 81.9%#1 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 17.0%#47 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 82.6%#20 | Scale AI SEAL |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
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 | 107 tok/s | 2.56 s | $0.250 | $2.00 | 400k | – |
| OpenAI | 100 tok/s | 2.20 s | $0.125 | $1.00 | 400k | – |
| Azure | 79 tok/s | 3.49 s | $0.250 | $2.00 | 400k | – |
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 $0.025 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0007 | $0.0006 | 6.6 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0042 | $0.0022 | 10.1 s |
| Code edit | 6,000 / 1,500 | $0.0045 | $0.0035 | 20.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.023 | $0.013 | 49.7 s |
| Structured extraction | 2,000 / 200 | $0.0009 | $0.0006 | 5.5 s |
More from OpenAI
See also
Questions people ask
- What are GPT-5 mini's benchmark scores?
- As of 24 Oct 2026, GPT-5 mini scores 56.5 ±7.6 on the BenchLeader Index, strongest in maths (79) and weakest in knowledge (40). That rests on 11 independent benchmark results, and every score on this page links to its publisher.
- How does GPT-5 mini rank?
- GPT-5 mini is #202 of 785 ranked configurations on 24 Oct 2026. 360 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-5 mini cost per 1M tokens?
- As of 24 Oct 2026, GPT-5 mini costs $0.250 per 1M input tokens and $2.00 per 1M output tokens, $0.688 blended (three input tokens per output token).
- When was GPT-5 mini released?
- OpenAI released GPT-5 mini on 7 Aug 2025. BenchLeader has results for it at 6 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GPT-5 mini: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/gpt-5-mini.