OpenAIGPT-5.4 Mini
GPT-5.4 Mini is an OpenAI proprietary reasoning model, released 17 Mar 2026. Its best configuration (xhigh reasoning effort) ranks #206 of 785 on the BenchLeader Index at 56.3 ±4.6, in the upper half. The ± is the point: 276 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 (72) and lowest in agents & tools (50). At $1.69 per million tokens blended it is pricier than most ranked models. Output speed of 241 tokens per second puts it in the fastest quarter, with a first answer in 112.9 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 1 Sept 2026.
- Blended price
- $1.69/M
- $0.750 in · $4.50 out
- Output speed
- 241 tok/s
- measured by Artificial Analysis
- First answer
- 113 s
- first token 1.00 s
- Context
- 400k
- Full answer
- 115 s
- median, reasoning included
- Cost per run
- $0.447
- one full Intelligence Index run
- Released
- 17 Mar 2026
- Overall index56
- Reasoning53
- Coding58
- Agents & tools50
- Maths51
- Knowledge54
- Instruction following72
- Multimodal56
- Long context64
- Composite56
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 Minixhighthis page | 17 Mar 2026 | 56.3 | #206 |
| GPT-5 mini | 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 |
|---|---|---|---|---|---|---|
| no reasoning | 44.0 | #552 | 176 tok/s | 0.65 s | $0.0017 | Agents & tools 50 · Coding 42 · Composite 42 · Instruction following 42 · Knowledge 41 · Long context 43 · Maths 35 · Multimodal 43 · Reasoning 44 |
| low | – | – | 97 tok/s | 1.00 s | $0.0017 | Maths 39 · Reasoning 31 |
| medium | 53.6 | #283 | 213 tok/s | 3.68 s | $0.0017 | Agents & tools 64 · Composite 52 · Instruction following 64 · Knowledge 52 · Long context 58 · Multimodal 54 · Reasoning 40 |
| high | 52.2 | #325 | 97 tok/s | 1.00 s | $0.0017 | Agents & tools 44 · Coding 52 · Human preference 63 · Knowledge 40 · Maths 62 · Multimodal 58 · Reasoning 51 |
| xhighbest | 56.3 | #206 | 241 tok/s | 113 s | $0.0017 | Agents & tools 50 · Coding 58 · Composite 56 · Instruction following 72 · Knowledge 54 · Long context 64 · Maths 51 · Multimodal 56 · Reasoning 53 |
| not stated | – | – | 97 tok/s | 1.00 s | $0.0017 | Coding 38 · Reasoning 36 |
How its index has moved
8 Sept 2026 to 23 Oct 2026- xhigh
- medium
- high
- no reasoning
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 | 64.1%#184 | – | – | 83.6%#100 | 86.9%#77 | – | Epoch AI Benchmarking Hub | |
| Arena Hard Prompts | – | – | – | 1469#82 | – | – | Arena (LMArena) | |
| LiveBench 1.0 Reasoningnot in index | – | – | – | – | 71.3%#63 | – | LiveBench | |
| GPQA Diamond (AA)not in index | 60.6%#371 | – | 82.3%#172 | – | 87.5%#97 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 5.9%#413 | – | 18.6%#228 | – | 28.1%#167 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | – | – | 83.1%#57 | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | – | – | – | 37.9%#109 | Kagi LLM Benchmark | |
| ARC-AGI-1 | – | 13.0%#231 | 40.8%#191 | 58.0%#165 | 63.7%#155 | – | ARC Prize | |
| ARC-AGI-2 | – | 1.1%#227 | 4.4%#183 | 13.2%#153 | 18.9%#144 | – | ARC Prize | |
| CritPt | 0.0%#315 | – | 2.9%#182 | – | 10.0%#121 | – | Artificial Analysis | |
| Chess Puzzlesnot in index | 2.0%#140 | – | – | 23.0%#82 | 17.0%#102 | – | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in index | 11.0%#111 | 8.0%#119 | 7.0%#123 | – | – | – | Epoch AI Benchmarking Hub | |
| LMCAnot in index | – | – | – | – | 40.8%#106 | – | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | – | – | – | 80.0%#132 | – | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | – | – | – | 61.2%#68 | ForecastBench |
Coding
| Benchmark | no reasoning | high | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| SciCode | – | – | 52.1%#88 | – | SciCode | |
| WeirdML | 37.9%#116 | 60.3%#50 | – | – | WeirdML | |
| FrontierCodeeffort not stated | – | – | – | 27.0%#34 | Cognition | |
| Arena Coding | – | 1495#88 | – | – | Arena (LMArena) | |
| Arena WebDev | – | 1397#93 | – | – | Arena (LMArena) | |
| LiveBench 1.0 Codingnot in index | – | – | 71.6%#59 | – | LiveBench | |
| SciCode (AA)not in index | – | – | 52.1%#88 | – | Artificial Analysis | |
| LiveCodeBench | – | – | 81.5%#59 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | – | 73.0%#51 | – | Vals AI | |
| Code Migrationnot in index | – | – | 12.9%#59 | – | Vals AI | |
| ProgramBenchnot in index | – | – | 0.0%#32 | – | Vals AI | |
| Vibe Code Bench v1.1not in index | – | – | 48.0%#63 | – | Vals AI | |
| ALE-Benchnot in index | – | 1188.6#44 | – | – | Epoch AI Benchmarking Hub | |
| Terminal-Bench 4.0 (AA)not in index | – | – | 2.0%#118 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | – | – | 59.2%#101 | – | Artificial Analysis |
Agents & tools
Maths
| Benchmark | no reasoning | low | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 17.2%#99 | 24.6%#81 | – | 51.2%#53 | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | 9.8%#65 | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 26.7%#212 | – | 87.2%#84 | 88.9%#76 | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | 21.0%#49 | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | – | 78.5%#62 | LiveBench | |
| AIME (Vals) | – | – | – | 95.6%#8 | Vals AI | |
| Arena Maths | – | – | 1439#93 | – | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | medium | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | 29.4%#68 | – | Epoch AI Benchmarking Hub | |
| LiveBench 1.0 Data Analysisnot in index | – | – | – | 70.8%#48 | LiveBench | |
| AA-Omniscience | -45.4#353 | -20.6#245 | – | -18.9#238 | Artificial Analysis | |
| MMLU-Pro | – | – | – | 84.5%#61 | Vals AI | |
| CorpFin | – | – | – | 60.9%#62 | Vals AI | |
| TaxEval | – | – | – | 71.2%#77 | Vals AI | |
| CaseLaw v2not in index | – | – | – | 51.7%#46 | Vals AI | |
| Excel Modeling Benchmarknot in index | – | – | – | 45.4%#56 | Vals AI | |
| AA-Omniscience: accuracynot in index | 25.6%#227 | 36.5%#149 | – | 37.5%#141 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 4.4%#531 | 10.2%#417 | – | 9.8%#425 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | medium | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | – | 71.0%#59 | LiveBench | |
| LiveBench Instruction Followingnot in index | – | – | – | 59.8%#58 | LiveBench | |
| IFBench | 38.8%#276 | 64.8%#105 | – | 73.3%#42 | Artificial Analysis | |
| Arena Instruction Followingnot in index | – | – | 1434#88 | – | Arena (LMArena) |
Human preference
| Benchmark | high | Source | Trend |
|---|---|---|---|
| Arena Text | 1447#83 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1402#105 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1466#71 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1447#93 | Arena (LMArena) |
Multimodal
| Benchmark | no reasoning | medium | high | xhigh | Source | Trend |
|---|---|---|---|---|---|---|
| Arena Vision | – | – | 1245#68 | – | Arena (LMArena) | |
| MMMU-Pro | 60.5%#209 | 71.2%#139 | – | 73.3%#124 | Artificial Analysis | |
| MMMU-Pro (Vals)not in index | – | – | – | 79.3%#48 | Vals AI | |
| MortgageTaxnot in index | – | – | – | 63.5%#50 | Vals AI | |
| SAGEnot in index | – | – | – | 50.8%#15 | Vals AI |
Long context
| Benchmark | no reasoning | medium | xhigh | Source | Trend |
|---|---|---|---|---|---|
| AA-LCR | 37.0%#370 | 67.0%#243 | 77.0%#138 | Artificial Analysis | |
| MLCRnot in index | – | – | 20.0%#47 | Artificial Analysis |
Composite
| Benchmark | no reasoning | medium | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | – | 148.7#65 | Epoch AI Benchmarking Hub | |
| LiveBench 1.0not in index | – | – | 66.4%#63 | – | LiveBench | |
| AA Intelligence Index v4.3.2 | 11.1#361 | 19.8#238 | 24.1#180 | – | Artificial Analysis | |
| Vals Multimodal Indexnot in index | – | – | 54.2%#21 | – | 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 | 107 tok/s | 1.51 s | $0.375 | $2.25 | 400k | – |
| OpenAI | 97 tok/s | 0.97 s | $0.750 | $4.50 | 400k | – |
| OpenAI | 97 tok/s | 0.88 s | $1.50 | $9.00 | 400k | – |
| Azure | 53 tok/s | 1.38 s | $0.750 | $4.50 | 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 $0.075 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0017 | $0.0014 | 1.9 min |
| Summarize a 30-page report | 12,000 / 600 | $0.012 | $0.0056 | 1.9 min |
| Code edit | 6,000 / 1,500 | $0.011 | $0.0082 | 2.0 min |
| Agentic coding session | 60,000 / 4,000 | $0.063 | $0.033 | 2.2 min |
| Structured extraction | 2,000 / 200 | $0.0024 | $0.0014 | 1.9 min |
More from OpenAI
See also
Questions people ask
- What are GPT-5.4 Mini's benchmark scores?
- As of 24 Oct 2026, GPT-5.4 Mini (xhigh effort) scores 56.3 ±4.6 on the BenchLeader Index, strongest in instruction following (72) and weakest in agents & tools (50). That rests on 22 independent benchmark results, and every score on this page links to its publisher.
- How does GPT-5.4 Mini rank?
- GPT-5.4 Mini (xhigh effort) is #206 of 785 ranked configurations on 24 Oct 2026. 276 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 Mini cost per 1M tokens?
- As of 24 Oct 2026, GPT-5.4 Mini costs $0.750 per 1M input tokens and $4.50 per 1M output tokens, $1.69 blended (three input tokens per output token).
- When was GPT-5.4 Mini released?
- OpenAI released GPT-5.4 Mini 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 Mini: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/gpt-5-4-mini.