OpenAIGPT-5.4 Nano
GPT-5.4 Nano is an OpenAI proprietary reasoning model, released 17 Mar 2026. Its best configuration (xhigh reasoning effort) ranks #259 of 785 on the BenchLeader Index at 54.2 ±6.4, in the upper half. The ± is the point: 352 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 (74) and lowest in reasoning (45). At $0.463 per million tokens blended it is cheaper than most ranked models. Output speed of 164 tokens per second puts it in the fastest quarter, with a first answer in 45.6 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 Mar 2026.stale: no new result in six months
- Blended price
- $0.463/M
- $0.200 in · $1.25 out
- Output speed
- 164 tok/s
- measured by Artificial Analysis
- First answer
- 46 s
- first token 0.95 s
- Context
- 400k
- Full answer
- 49 s
- median, reasoning included
- Cost per run
- $0.183
- one full Intelligence Index run
- Released
- 17 Mar 2026
- Overall index54
- Reasoning45
- Coding50
- Agents & tools61
- Knowledge48
- Instruction following74
- Multimodal48
- Long context63
- Composite53
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 Nanoxhighthis page | 17 Mar 2026 | 54.2 | #259 |
| GPT-5 nano | 7 Aug 2025 | 55.9 | #220 |
| 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 |
|---|---|---|---|---|---|---|
| no reasoning | 42.1 | #607 | 163 tok/s | 0.64 s | $0.0005 | Agents & tools 55 · Coding 42 · Composite 43 · Instruction following 36 · Knowledge 45 · Long context 39 · Maths 33 · Multimodal 26 · Reasoning 41 |
| low | 44.5 | #536 | 91 tok/s | 0.95 s | $0.0005 | Maths 42 · Reasoning 40 |
| medium | 52.4 | #319 | 161 tok/s | 3.32 s | $0.0005 | Agents & tools 63 · Composite 52 · Instruction following 64 · Knowledge 53 · Long context 59 · Multimodal 42 · Reasoning 40 |
| high | 48.4 | #427 | 91 tok/s | 0.95 s | $0.0005 | Agents & tools 32 · Coding 58 · Human preference 58 · Knowledge 41 · Maths 51 · Multimodal 51 · Reasoning 46 |
| xhighbest | 54.2 | #259 | 164 tok/s | 46 s | $0.0005 | Agents & tools 61 · Coding 50 · Composite 53 · Instruction following 74 · Knowledge 48 · Long context 63 · Multimodal 48 · Reasoning 45 |
| not stated | – | – | 91 tok/s | 0.95 s | $0.0005 | Reasoning 38 |
How its index has moved
8 Sept 2026 to 23 Oct 2026- xhigh
- medium
- high
- 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 | no reasoning | low | medium | high | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| GPQA Diamond | 55.6%#204 | 72.2%#162 | – | 78.5%#134 | – | – | Epoch AI Benchmarking Hub | |
| Arena Hard Prompts | – | – | – | 1421#144 | – | – | Arena (LMArena) | |
| LiveBench 1.0 Reasoningnot in index | – | – | – | – | 81.1%#51 | – | LiveBench | |
| GPQA Diamond (AA)not in index | 55.8%#401 | – | 76.1%#240 | – | 81.7%#178 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 4.1%#544 | – | 15.9%#245 | – | 28.3%#165 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | – | 77.5%#73 | – | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | – | – | – | 39.7%#107 | Kagi LLM Benchmark | |
| ARC-AGI-1 | – | 18.3%#223 | 33.0%#205 | 38.2%#195 | 51.5%#178 | – | ARC Prize | |
| ARC-AGI-2 | – | 1.5%#214 | 1.9%#205 | 3.6%#191 | 5.7%#173 | – | ARC Prize | |
| CritPt | 0.0%#315 | – | 5.1%#154 | – | 9.3%#127 | – | Artificial Analysis | |
| Chess Puzzlesnot in index | 2.0%#140 | 20.0%#92 | – | 36.0%#49 | 45.0%#29 | – | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in index | 9.0%#115 | 3.0%#138 | 6.0%#130 | 5.0%#132 | – | – | Epoch AI Benchmarking Hub | |
| LMCAnot in index | – | – | – | – | 36.9%#122 | – | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | – | – | – | 80.3%#130 | – | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | – | – | – | 58.6%#79 | ForecastBench |
Coding
| Benchmark | no reasoning | high | xhigh | Source | Trend |
|---|---|---|---|---|---|
| SciCode | – | – | 47.2%#126 | SciCode | |
| WeirdML | 38.0%#115 | 49.2%#72 | – | WeirdML | |
| Arena Coding | – | 1460#129 | – | Arena (LMArena) | |
| LiveBench 1.0 Codingnot in index | – | – | 70.8%#60 | LiveBench | |
| SciCode (AA)not in index | – | – | 47.2%#134 | Artificial Analysis | |
| LiveCodeBench | – | 84.0%#40 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | 69.8%#61 | – | Vals AI | |
| Code Migrationnot in index | – | 14.5%#55 | – | Vals AI | |
| Vibe Code Bench v1.1not in index | – | 26.1%#75 | – | Vals AI | |
| ALE-Benchnot in index | – | 1004.5#58 | – | Epoch AI Benchmarking Hub | |
| Terminal-Bench 4.0 (AA)not in index | – | – | 0.5%#139 | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | – | – | 60.7%#98 | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | medium | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| LiveBench Agentic Codingnot in index | – | – | – | 46.8%#50 | LiveBench | |
| Terminal-Bench Hard | 24.2%#144 | 33.3%#92 | – | 42.4%#46 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 34.8%#237 | 52.6%#193 | – | 76.0%#134 | Artificial Analysis | |
| Terminal-Bench 2.1 (Vals) | – | – | 41.6%#66 | – | Vals AI | |
| τ³-Banking (AA) | – | – | – | 27.4%#76 | Artificial Analysis | |
| ITBench SRE (AA)not in index | 12.9%#42 | – | – | 24.4%#37 | Artificial Analysis | |
| APEX-Agents (AA)not in index | – | – | – | 24.9%#18 | Artificial Analysis | |
| Finance Agent v2not in index | – | – | 38.2%#62 | – | Vals AI | |
| Harvey's Legal Agent Benchmark v1not in index | – | – | 0.0%#61 | – | Vals AI | |
| Legal Research Benchnot in index | – | – | 6.3%#74 | – | Vals AI | |
| Tax Agent Benchnot in index | – | – | 5.0%#67 | – | Vals AI | |
| Terminal-Bench 2.0 (Vals)not in index | – | – | 39.9%#34 | – | Vals AI | |
| AutomationBenchnot in index | – | – | – | 6.0%#162 | Artificial Analysis | |
| GDP.pdf (AA)not in index | – | – | – | 7.8%#146 | Artificial Analysis | |
| EnterpriseOps-Gymnot in index | – | – | – | 31.5%#39 | Artificial Analysis |
Maths
| Benchmark | no reasoning | low | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 4.6%#112 | 20.4%#87 | 44.9%#57 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | 12.2%#62 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 46.7%#187 | 68.9%#141 | 87.8%#83 | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | 5.0%#71 | – | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | – | 91.0%#33 | LiveBench | |
| AIME (Vals) | – | – | 88.8%#28 | – | Vals AI | |
| Arena Maths | – | – | 1424#113 | – | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | medium | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | 11.7%#83 | – | Epoch AI Benchmarking Hub | |
| LiveBench 1.0 Data Analysisnot in index | – | – | – | 67.6%#55 | LiveBench | |
| AA-Omniscience | -37.5#312 | -18.0#237 | – | -29.5#277 | Artificial Analysis | |
| MMLU-Pro | – | – | 77.2%#101 | – | Vals AI | |
| LegalBench | – | – | 77.9%#101 | – | Vals AI | |
| CorpFin | – | – | 61.2%#56 | – | Vals AI | |
| TaxEval | – | – | 67.4%#103 | – | Vals AI | |
| CaseLaw v2not in index | – | – | 51.9%#45 | – | Vals AI | |
| Excel Modeling Benchmarknot in index | – | – | 44.8%#57 | – | Vals AI | |
| MedCodenot in index | – | – | 41.0%#57 | – | Vals AI | |
| MedScribenot in index | – | – | 77.1%#70 | – | Vals AI | |
| AA-Omniscience: accuracynot in index | 14.9%#426 | 21.9%#280 | – | 25.7%#226 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 38.5%#169 | 48.9%#122 | – | 25.8%#236 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | medium | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | – | 62.5%#65 | LiveBench | |
| LiveBench Instruction Followingnot in index | – | – | – | 67.2%#37 | LiveBench | |
| IFBench | 32.7%#338 | 64.4%#108 | – | 75.9%#24 | Artificial Analysis | |
| Arena Instruction Followingnot in index | – | – | 1387#151 | – | Arena (LMArena) |
Human preference
| Benchmark | high | Source | Trend |
|---|---|---|---|
| Arena Text | 1401#149 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1337#180 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1413#138 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1403#157 | Arena (LMArena) |
Multimodal
| Benchmark | no reasoning | medium | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| Arena Vision | – | – | 1196#93 | – | Arena (LMArena) | |
| MMMU-Pro | 43.8%#258 | 59.5%#215 | – | 65.4%#175 | Artificial Analysis | |
| MMMU-Pro (Vals)not in index | – | – | 73.6%#57 | – | Vals AI | |
| MortgageTaxnot in index | – | – | 59.1%#65 | – | Vals AI | |
| SAGEnot in index | – | – | 38.1%#63 | – | Vals AI |
Long context
| Benchmark | no reasoning | medium | xhigh | Source | Trend |
|---|---|---|---|---|---|
| AA-LCR | 29.7%#396 | 67.3%#240 | 76.7%#140 | Artificial Analysis | |
| MLCRnot in index | – | – | 16.1%#69 | Artificial Analysis |
Composite
| Benchmark | no reasoning | medium | high | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | – | – | 146.0#78 | Epoch AI Benchmarking Hub | |
| LiveBench 1.0not in index | – | – | – | 69.6%#55 | – | LiveBench | |
| AA Intelligence Index v4.3.2 | 11.7#347 | 20.0#236 | – | 20.7#224 | – | Artificial Analysis | |
| Vals Multimodal Indexnot in index | – | – | 47.6%#26 | – | – | Vals AI |
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 | 91 tok/s | 0.87 s | $0.200 | $1.25 | 400k | – |
| Azure | 65 tok/s | 1.75 s | $0.200 | $1.25 | 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 $0.020 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0005 | $0.0004 | 47.4 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0032 | $0.0015 | 49.2 s |
| Code edit | 6,000 / 1,500 | $0.0031 | $0.0023 | 54.7 s |
| Agentic coding session | 60,000 / 4,000 | $0.017 | $0.0089 | 1.2 min |
| Structured extraction | 2,000 / 200 | $0.0007 | $0.0004 | 46.8 s |
More from OpenAI
See also
Questions people ask
- What are GPT-5.4 Nano's benchmark scores?
- As of 24 Oct 2026, GPT-5.4 Nano (xhigh effort) scores 54.2 ±6.4 on the BenchLeader Index, strongest in instruction following (74) and weakest in reasoning (45). That rests on 11 independent benchmark results, and every score on this page links to its publisher.
- How does GPT-5.4 Nano rank?
- GPT-5.4 Nano (xhigh effort) is #259 of 785 ranked configurations on 24 Oct 2026. 352 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-5.4 Nano cost per 1M tokens?
- As of 24 Oct 2026, GPT-5.4 Nano costs $0.200 per 1M input tokens and $1.25 per 1M output tokens, $0.463 blended (three input tokens per output token).
- When was GPT-5.4 Nano released?
- OpenAI released GPT-5.4 Nano on 17 Mar 2026. BenchLeader has results for it at 6 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GPT-5.4 Nano: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/gpt-5-4-nano.