OpenAIGPT-5.4 mini
Best configuration ranks #191 of 610 on the BenchLeader Index at 54.8 ±8.3. Last measured 17 Mar 2026. Released 17 Mar 2026.
- Blended price
- $1.69/M
- $0.750 in · $4.50 out
- Output speed
- 218 tok/s
- First answer
- 139 s
- first token 1.31 s
- Context
- 400k
- Overall index55
- Reasoning36
- Coding42
- Agents & tools65
- Knowledge55
- Instruction following71
- Multimodal58
- Long context65
- Composite60
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. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | 44.9 | #402 | 145 tok/s | 0.71 s | $0.0017 | Agents & tools 49 · Coding 42 · Composite 43 · Instruction following 41 · Knowledge 43 · Long context 44 · Maths 36 · Multimodal 45 · Reasoning 48 |
| low | – | – | 218 tok/s | 139 s | $0.0017 | Maths 41 · Reasoning 32 |
| medium | 53.9 | #211 | 173 tok/s | 6.27 s | $0.0017 | Agents & tools 63 · Composite 54 · Instruction following 64 · Knowledge 54 · Long context 60 · Multimodal 56 · Reasoning 38 |
| high | 52.9 | #236 | 218 tok/s | 139 s | $0.0017 | Agents & tools 44 · Coding 53 · Human preference 64 · Knowledge 41 · Maths 62 · Multimodal 59 · Reasoning 53 |
| xhigh | 46.0 | #372 | 218 tok/s | 139 s | $0.0017 | Agents & tools 33 · Coding 54 · Composite 25 · Knowledge 55 · Maths 53 · Reasoning 53 |
| defaultbest | 54.8 | #191 | 218 tok/s | 139 s | $0.0017 | Agents & tools 65 · Coding 42 · Composite 60 · Instruction following 71 · Knowledge 55 · Long context 65 · Multimodal 58 · Reasoning 36 |
Benchmark results
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 | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| GPQA Diamond | 64.1%#170 | – | – | 83.6%#88 | 86.9%#65 | – | Epoch AI Benchmarking Hub | |
| LMArena Hard Prompts | – | – | – | 1470#70 | – | – | LMArena | |
| LiveBench Reasoningnot in index | – | – | – | – | 71.3%#49 | – | LiveBench | |
| GPQA Diamond (AA)not in index | 60.6%#352 | – | 82.3%#168 | – | – | 87.5%#96 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 5.9%#338 | – | 18.6%#171 | – | – | 28.1%#119 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | – | – | 83.1%#57 | – | Vals AI | |
| Kagi LLM Benchmark | – | – | – | – | – | 37.9%#108 | Kagi LLM Benchmark | |
| ARC-AGI-1 | – | 13.0%#165 | 40.8%#131 | 58.0%#111 | 63.7%#104 | – | ARC Prize | |
| ARC-AGI-2 | – | 1.1%#161 | 4.4%#125 | 13.2%#101 | 18.9%#94 | – | ARC Prize |
Coding
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SciCode | – | – | – | – | 49.9%#62 | – | SciCode | |
| WeirdML | 37.9%#114 | – | – | 60.3%#47 | – | – | WeirdML | |
| FrontierCode | – | – | – | – | – | 27.0%#19 | Cognition | |
| LMArena Coding | – | – | – | 1496#70 | – | – | LMArena | |
| LMArena WebDev | – | – | – | 1397#72 | – | – | LMArena | |
| LiveBench Codingnot in index | – | – | – | – | 71.6%#45 | – | LiveBench | |
| SciCode (AA)not in index | – | – | – | – | – | 52.1%#55 | Artificial Analysis | |
| LiveCodeBench | – | – | – | – | 81.5%#59 | – | Vals AI | |
| IOI | – | – | – | – | 6.4%#41 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | – | – | – | 73.0%#51 | – | Vals AI |
Agents & tools
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| APEX-Agents | – | – | – | – | – | 24.6%#30 | Mercor | |
| LiveBench Agentic Codingnot in index | – | – | – | – | 41.7%#46 | – | LiveBench | |
| Terminal-Bench Hard | 18.2%#166 | – | 34.1%#87 | – | – | 52.3%#19 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 23.4%#291 | – | 36.5%#223 | – | – | 83.3%#106 | Artificial Analysis | |
| Terminal-Bench 2.1 (Vals) | – | – | – | 54.7%#40 | – | – | Vals AI | |
| MCP Atlas | – | – | – | – | 56.7%#27 | – | Scale AI SEAL |
Maths
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 17.2%#84 | 24.6%#66 | – | – | 51.2%#38 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | – | 9.8%#50 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 26.7%#193 | – | – | 87.2%#69 | 88.9%#61 | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | – | 21.0%#34 | – | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | – | – | 78.5%#48 | – | LiveBench | |
| AIME (Vals) | – | – | – | – | 95.6%#8 | – | Vals AI |
Knowledge
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | – | 29.4%#58 | – | – | Epoch AI Benchmarking Hub | |
| LiveBench Data Analysisnot in index | – | – | – | – | 70.8%#36 | – | LiveBench | |
| AA-Omniscience | -45.4#286 | – | -20.6#184 | – | – | -18.9#178 | Artificial Analysis | |
| MMLU-Pro | – | – | – | – | 84.5%#61 | – | Vals AI | |
| CorpFin | – | – | – | – | 60.9%#62 | – | Vals AI | |
| TaxEval | – | – | – | – | 71.2%#78 | – | Vals AI |
Instruction following
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | – | – | 71.0%#48 | – | LiveBench | |
| IFBench | 38.8%#264 | – | 64.8%#104 | – | – | 73.3%#42 | Artificial Analysis |
Human preference
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LMArena Text | – | – | – | 1448#70 | – | – | LMArena |
Multimodal
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LMArena Vision | – | – | – | 1246#53 | – | – | LMArena | |
| MMMU-Pro | 60.5%#175 | – | 71.2%#109 | – | – | 73.3%#96 | Artificial Analysis |
Long context
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| AA-LCR | 37.0%#299 | – | 67.0%#185 | – | – | 77.0%#95 | Artificial Analysis |
Composite
| Benchmark | no reasoning | low | medium | high | xhigh | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | – | – | – | 149.0#50 | Epoch AI Benchmarking Hub | |
| LiveBench | – | – | – | – | 66.4%#49 | – | LiveBench | |
| AA Intelligence Index | 11.1#293 | – | 19.8#176 | – | – | 24.6#123 | Artificial Analysis | |
| Vals Indexnot in index | – | – | – | – | 39.6#36 | – | 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 Fast | 78 tok/s | 1.47 s | $1.50 | $9.00 | 400k | – |
| OpenAI | 59 tok/s | 1.16 s | $0.750 | $4.50 | 400k | – |
| Azure (US) | 43 tok/s | 1.68 s | $0.825 | $4.95 | 400k | – |
| Azure | 39 tok/s | 1.08 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 | 2.3 min |
| Summarise a 30-page report | 12,000 / 600 | $0.012 | $0.0056 | 2.4 min |
| Code edit | 6,000 / 1,500 | $0.011 | $0.0082 | 2.4 min |
| Agentic coding session | 60,000 / 4,000 | $0.063 | $0.033 | 2.6 min |
| Structured extraction | 2,000 / 200 | $0.0024 | $0.0014 | 2.3 min |
See also
Data as of 9 Sept 2026. Compare these configurations.