OpenAIReasoning modelGPT-5.4 Pro
GPT-5.4 Pro is an OpenAI proprietary reasoning model, released 4 Mar 2026. Its best configuration ranks #60 of 751 on the BenchLeader Index at 64.0 ±3.0, in the upper half. The ± is the point: 121 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 knowledge (74) and lowest in instruction following (67). At $67.50 per million tokens blended it is among the most expensive ranked models. Output speed of 2 tokens per second puts it in the slowest quarter, with a first token in 6.1 s. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 23 Mar 2026.stale: no new result in six months
- Blended price
- $67.50/M
- $30.00 in · $180.00 out
- Output speed
- 2 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 6.09 s
- Context
- 1.1M
- Full answer
- –
- Cost per run
- –
- Released
- 4 Mar 2026
- Overall index64
- Reasoning73
- Knowledge74
- Instruction following67
- Multimodal70
Versions
OpenAI has shipped 4 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-5.5 Proxhigh | 23 Apr 2026 | 65.6 | #39 |
| GPT-5.4 Prothis page | 4 Mar 2026 | 64.0 | #60 |
| GPT-5.2 Proxhigh | 11 Dec 2025 | – | – |
| GPT-5 Pro | 7 Aug 2025 | 60.2 | #108 |
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 | – | – | 2 tok/s | 6.09 s | $0.066 | Coding 55 |
| xhigh | 62.1 | #86 | 2 tok/s | 6.09 s | $0.066 | Knowledge 54 · Maths 71 · Reasoning 75 |
| not statedbest | 64.0 | #60 | 2 tok/s | 6.09 s | $0.066 | Instruction following 67 · Knowledge 74 · Multimodal 70 · Reasoning 73 |
How its index has moved
8 Sept 2026 to 3 Oct 2026- not stated
- xhigh
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
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 | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|
| GPQA Diamond | – | 94.6%#6 | – | Epoch AI Benchmarking Hub | |
| Humanity's Last Exam | – | – | 44.3%#5 | Scale AI / CAIS | |
| SimpleBench | – | – | 74.1%#10 | SimpleBench | |
| ARC-AGI-1 | – | 94.5%#35 | – | ARC Prize | |
| ARC-AGI-2 | – | 83.3%#34 | – | ARC Prize | |
| CritPt | – | 30.0%#15 | – | Artificial Analysis | |
| Chess Puzzlesnot in index | – | 58.6%#6 | – | Epoch AI Benchmarking Hub |
Maths
| Benchmark | no reasoning | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | – | 82.5%#13 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | 58.5%#21 | – | Epoch AI Benchmarking Hub | |
| MathArena Apex | – | 69.8%#3 | – | MathArena |
Knowledge
| Benchmark | no reasoning | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|
| SimpleQA Verified | – | 46.3%#35 | – | Epoch AI Benchmarking Hub | |
| MultiNRC | – | – | 62.3%#4 | Scale AI SEAL |
Instruction following
| Benchmark | no reasoning | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|
| MultiChallenge | – | – | 69.2%#4 | Scale AI SEAL | |
| TutorBench | – | – | 56.6%#2 | Scale AI SEAL |
Multimodal
| Benchmark | no reasoning | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|
| VISTA | – | – | 53.9%#2 | Scale AI SEAL |
Composite
| Benchmark | no reasoning | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | 159.1#11 | Epoch AI Benchmarking Hub |
Safety & honesty
| Benchmark | no reasoning | xhigh | not stated | Source | Trend |
|---|---|---|---|---|---|
| FORTRESSnot in index | – | – | 14.8%#53 | Scale AI SEAL | |
| MASKnot in index | – | – | 91.7%#7 | 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 | 6 tok/s | 6.09 s | $30.00 | $180.00 | 1.1M | – |
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. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.066 | – | 2.6 min |
| Summarise a 30-page report | 12,000 / 600 | $0.468 | – | 5.1 min |
| Code edit | 6,000 / 1,500 | $0.450 | – | 12.6 min |
| Agentic coding session | 60,000 / 4,000 | $2.52 | – | 33.4 min |
| Structured extraction | 2,000 / 200 | $0.096 | – | 1.8 min |
See also
Data as of 4 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GPT-5.4 Pro: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/gpt-5-4-pro, data as of 4 Oct 2026.