OpenAIGPT-5 nano
GPT-5 nano is an OpenAI proprietary reasoning model, released 7 Aug 2025. Its best configuration ranks #219 of 785 on the BenchLeader Index at 55.9 ±9.5, in the upper half. The ± is the point: 451 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 instruction following (66) and lowest in agents & tools (48). At $0.138 per million tokens blended it is among the cheapest fifth of ranked models. Output speed of 82 tokens per second puts it slower than most, with a first token in 1.3 s. It has been measured at 5 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 24 Nov 2025.stale: no new result in six months
- Blended price
- $0.138/M
- $0.050 in · $0.400 out
- Output speed
- 82 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 1.25 s
- Context
- 400k
- Full answer
- –
- Cost per run
- –
- Released
- 7 Aug 2025
- Overall index56
- Reasoning57
- Agents & tools48
- Maths64
- Instruction following66
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 Nanoxhigh | 17 Mar 2026 | 54.2 | #259 |
| GPT-5 nanothis page | 7 Aug 2025 | 55.9 | #219 |
| GPT-4.1 nano | 14 Apr 2025 | 38.0 | #720 |
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 | 35.8 | #759 | 124 tok/s | 0.89 s | $0.0001 | Agents & tools 40 · Composite 37 · Instruction following 36 · Knowledge 33 · Long context 34 · Maths 31 · Multimodal 14 · Reasoning 34 |
| low | 39.3 | #689 | 82 tok/s | 1.25 s | $0.0001 | Coding 34 · Maths 34 · Reasoning 35 |
| medium | 45.9 | #503 | 138 tok/s | 31 s | $0.0001 | Agents & tools 34 · Coding 35 · Composite 43 · Instruction following 65 · Knowledge 50 · Long context 36 · Maths 62 · Multimodal 41 · Reasoning 40 |
| high | 47.4 | #455 | 117 tok/s | 84 s | $0.0001 | Agents & tools 44 · Coding 48 · Composite 44 · Human preference 50 · Instruction following 67 · Knowledge 38 · Long context 47 · Maths 48 · Multimodal 45 · Reasoning 41 |
| not statedbest | 55.9 | #219 | 82 tok/s | 1.25 s | $0.0001 | Agents & tools 48 · Instruction following 66 · Maths 64 · Reasoning 57 |
How its index has moved
8 Sept 2026 to 23 Oct 2026- not stated
- high
- medium
- low
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 | 48.5%#221 | 57.6%#195 | 67.4%#175 | 69.4%#172 | – | Epoch AI Benchmarking Hub | |
| Arena Hard Prompts | – | – | – | 1356#210 | – | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 42.8%#462 | – | 67.0%#326 | 67.6%#319 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 4.0%#552 | – | 8.7%#345 | 9.5%#334 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | – | 63.4%#101 | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | – | – | 62.2%#46 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | – | – | – | 67.9%#13 | HELM Capabilities | |
| ARC-AGI-1 | 1.5%#254 | 4.0%#250 | 20.7%#221 | 16.7%#225 | – | ARC Prize | |
| ARC-AGI-2 | 0.0%#242 | 0.0%#242 | 0.9%#229 | 2.6%#197 | – | ARC Prize | |
| CritPt | 0.0%#315 | – | 0.0%#315 | 0.0%#315 | – | Artificial Analysis | |
| Chess Puzzlesnot in index | 2.0%#140 | 15.0%#107 | – | 22.0%#84 | – | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in index | 5.0%#132 | 9.0%#115 | 9.0%#115 | 8.0%#119 | – | Epoch AI Benchmarking Hub | |
| LMCAnot in index | – | – | – | 7.9%#184 | – | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | – | – | 62.7%#174 | – | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | – | – | 59.0%#75 | ForecastBench |
Coding
| Benchmark | low | medium | high | Source | Trend |
|---|---|---|---|---|---|
| WeirdML | 25.9%#134 | – | 38.1%#113 | WeirdML | |
| Arena Coding | – | – | 1385#210 | Arena (LMArena) | |
| LiveCodeBench | – | – | 70.2%#86 | Vals AI | |
| SWE-bench Verified (bash only) | – | 34.8%#37 | – | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | – | 34.8%#50 | – | SWE-bench | |
| ALE-Benchnot in index | – | – | 718.7#84 | Epoch AI Benchmarking Hub |
Agents & tools
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Terminal-Bench | – | 11.5%#81 | – | 21.8%#69 | Terminal-Bench | |
| Terminal-Bench Hard | 6.8%#249 | 17.4%#178 | 12.1%#210 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 25.7%#290 | 30.4%#258 | 36.5%#230 | – | Artificial Analysis | |
| BFCL Overalleffort not stated | – | – | – | 51.5%#22 | Berkeley Function Calling Leaderboard |
Maths
| Benchmark | minimal | low | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 1.8%#114 | 6.0%#109 | – | 20.0%#89 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | 2.4%#68 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 35.6%#203 | 46.7%#187 | 74.2%#128 | 81.1%#113 | – | Epoch AI Benchmarking Hub | |
| MATH Level 5 | – | – | 95.2%#13 | 94.9%#15 | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | 12.0%#63 | – | Vals AI | |
| AIME (Vals) | – | – | – | 81.2%#45 | – | Vals AI | |
| MGSM | – | – | – | 89.3%#44 | – | Vals AI | |
| Omni-MATH (HELM)effort not stated | – | – | – | – | 54.7%#16 | HELM Capabilities | |
| MATH 500not in index | – | – | – | 93.8%#15 | – | Vals AI | |
| Arena Maths | – | – | – | 1347#202 | – | Arena (LMArena) |
Knowledge
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | 11.7%#83 | – | Epoch AI Benchmarking Hub | |
| AA-Omniscience | -64.1#491 | -25.8#262 | -28.7#275 | – | Artificial Analysis | |
| MMLU-Pro | – | – | 76.1%#102 | – | Vals AI | |
| LegalBench | – | – | 50.1%#141 | – | Vals AI | |
| TaxEval | – | – | 67.4%#104 | – | Vals AI | |
| MedQA | – | – | 93.3%#23 | – | Vals AI | |
| MMLU-Pro (HELM)not in indexeffort not stated | – | – | – | 77.8%#26 | HELM Capabilities | |
| CaseLaw v2not in index | – | – | 52.6%#42 | – | Vals AI | |
| MedCodenot in index | – | – | 30.4%#92 | – | Vals AI | |
| MedScribenot in index | – | – | 72.9%#82 | – | Vals AI | |
| AA-Omniscience: accuracynot in index | 13.1%#462 | 17.6%#363 | 19.1%#319 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 11.2%#392 | 47.4%#133 | 41.0%#156 | – | Artificial Analysis |
Instruction following
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| IFBench | 32.5%#341 | 65.9%#101 | 67.5%#85 | – | Artificial Analysis | |
| IFEval (HELM)effort not stated | – | – | – | 93.2%#4 | HELM Capabilities | |
| WildBench (HELM)effort not stated | – | – | – | 80.6%#30 | HELM Capabilities | |
| Arena Instruction Followingnot in index | – | – | 1327#210 | – | Arena (LMArena) |
Human preference
| Benchmark | high | Source | Trend |
|---|---|---|---|
| Arena Text | 1338#215 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1247#266 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1325#219 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1333#225 | Arena (LMArena) |
Multimodal
| Benchmark | minimal | medium | high | Source | Trend |
|---|---|---|---|---|---|
| Arena Vision | – | – | 1160#112 | Arena (LMArena) | |
| MMMU-Pro | 31.8%#273 | 58.2%#219 | 61.0%#208 | Artificial Analysis | |
| MMMU-Pro (Vals)not in index | – | – | 70.9%#67 | Vals AI | |
| MortgageTaxnot in index | – | – | 53.6%#76 | Vals AI | |
| SAGEnot in index | – | – | 30.4%#80 | Vals AI | |
| VPCTnot in index | – | 35.4%#25 | 37.2%#23 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | minimal | medium | high | Source | Trend |
|---|---|---|---|---|---|
| Fiction.LiveBench 120k | – | 21.9%#33 | – | Fiction.live | |
| AA-LCR | 20.0%#436 | 43.7%#344 | 45.0%#338 | Artificial Analysis |
Composite
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | – | 139.2#116 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 7.1#495 | 12.5#328 | 13.0#319 | – | Artificial Analysis | |
| HELM Capabilities meannot in indexeffort not stated | – | – | – | 74.8%#14 | HELM Capabilities |
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 | 160 tok/s | 1.69 s | $0.025 | $0.200 | 400k | – |
| OpenAI | 97 tok/s | 1.24 s | $0.050 | $0.400 | 400k | – |
| Azure | 63 tok/s | 1.03 s | $0.050 | $0.400 | 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 $0.005 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0001 | $0.0001 | 4.9 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0008 | $0.0004 | 8.6 s |
| Code edit | 6,000 / 1,500 | $0.0009 | $0.0007 | 19.5 s |
| Agentic coding session | 60,000 / 4,000 | $0.0046 | $0.0026 | 50.0 s |
| Structured extraction | 2,000 / 200 | $0.0002 | $0.0001 | 3.7 s |
More from OpenAI
See also
Questions people ask
- What are GPT-5 nano's benchmark scores?
- As of 25 Oct 2026, GPT-5 nano scores 55.9 ±9.5 on the BenchLeader Index, strongest in instruction following (66) and weakest in agents & tools (48). That rests on 6 independent benchmark results, and every score on this page links to its publisher.
- How does GPT-5 nano rank?
- GPT-5 nano is #219 of 785 ranked configurations on 25 Oct 2026. 451 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 nano cost per 1M tokens?
- As of 25 Oct 2026, GPT-5 nano costs $0.050 per 1M input tokens and $0.400 per 1M output tokens, $0.138 blended (three input tokens per output token).
- When was GPT-5 nano released?
- OpenAI released GPT-5 nano on 7 Aug 2025. BenchLeader has results for it at 5 reasoning-effort settings.
Data as of 25 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GPT-5 nano: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/gpt-5-nano.