OpenAIGPT-5
GPT-5 is an OpenAI proprietary reasoning model, released 7 Aug 2025. Its best configuration (high reasoning effort) ranks #165 of 785 on the BenchLeader Index at 57.9 ±2.7, in the upper half. The ± is the point: 231 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 reasoning (53). At $3.44 per million tokens blended it is among the most expensive ranked models. Output speed of 91 tokens per second puts it faster than most, with a first answer in 70.2 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 22 Oct 2026.
- Blended price
- $3.44/M
- $1.25 in · $10.00 out
- Output speed
- 91 tok/s
- measured by Artificial Analysis
- First answer
- 70 s
- first token 4.66 s
- Context
- 400k
- Full answer
- 76 s
- median, reasoning included
- Cost per run
- –
- Released
- 7 Aug 2025
- Overall index58
- Reasoning53
- Coding56
- Agents & tools54
- Maths57
- Knowledge58
- Instruction following72
- Human preference62
- Multimodal55
- Long context64
- Composite55
Versions
OpenAI has shipped 13 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-5.6 | 9 Jul 2026 | – | – |
| GPT-5.5xhigh | 23 Apr 2026 | 66.1 | #35 |
| GPT-5.4xhigh | 5 Mar 2026 | 64.7 | #45 |
| GPT-5.2xhigh | 11 Dec 2025 | 60.7 | #104 |
| GPT-5.1high | 12 Nov 2025 | 57.7 | #170 |
| GPT-5highthis page | 7 Aug 2025 | 57.9 | #165 |
| GPT-4.1 | 14 Apr 2025 | 48.3 | #429 |
| GPT-4.5 | 27 Feb 2025 | 49.7 | #388 |
| GPT 4 0125 | 25 Jan 2024 | – | – |
| GPT-4 1106 | 6 Nov 2023 | – | – |
| GPT 4 0613 | 13 Jun 2023 | 39.7 | #677 |
| GPT-4 | 14 Mar 2023 | 42.4 | #596 |
| GPT-3.5 | – | – | – |
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 | 45.3 | #517 | 92 tok/s | 1.06 s | $0.0035 | Agents & tools 50 · Composite 42 · Instruction following 48 · Knowledge 46 · Maths 38 · Multimodal 45 · Reasoning 39 |
| low | 52.5 | #316 | 75 tok/s | 6.56 s | $0.0035 | Agents & tools 57 · Composite 53 · Instruction following 66 · Knowledge 57 · Maths 46 · Multimodal 57 · Reasoning 39 |
| medium | 57.5 | #175 | 87 tok/s | 23 s | $0.0035 | Agents & tools 57 · Coding 51 · Composite 55 · Instruction following 69 · Knowledge 57 · Long context 70 · Maths 66 · Multimodal 57 · Reasoning 46 |
| highbest | 57.9 | #165 | 91 tok/s | 70 s | $0.0035 | Agents & tools 54 · Coding 56 · Composite 55 · Human preference 62 · Instruction following 72 · Knowledge 58 · Long context 64 · Maths 57 · Multimodal 55 · Reasoning 53 |
| thinking | – | – | 66 tok/s | 4.66 s | $0.0035 | Instruction following 56 |
| not stated | 56.2 | #210 | 66 tok/s | 4.66 s | $0.0035 | Agents & tools 48 · Coding 60 · Composite 41 · Human preference 61 · Instruction following 57 · Knowledge 62 · Long context 57 · Maths 66 · Multimodal 59 · Reasoning 59 |
How its index has moved
8 Sept 2026 to 23 Oct 2026- high
- medium
- not stated
- 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 | 71.7%#163 | – | 85.3%#88 | 86.2%#82 | – | Epoch AI Benchmarking Hub | |
| Humanity's Last Exam | – | – | – | 25.3%#15 | 25.3%#15 | Scale AI / CAIS | |
| SimpleBench | – | – | – | 56.7%#36 | – | SimpleBench | |
| Arena Hard Prompts | – | – | – | 1448#111 | 1448#110 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 67.3%#321 | 80.8%#189 | 84.2%#148 | 85.3%#128 | 68.6%#311 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 6.0%#408 | 19.6%#219 | 25.4%#183 | 28.5%#162 | 6.6%#384 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | – | 85.6%#44 | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | – | – | 72.7%#21 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | – | – | – | 79.1%#2 | HELM Capabilities | |
| ARC-AGI-1 | 6.0%#240 | 44.0%#188 | 56.2%#171 | 65.7%#152 | – | ARC Prize | |
| ARC-AGI-2 | 0.0%#242 | 1.9%#205 | 7.5%#166 | 9.9%#160 | – | ARC Prize | |
| CritPt | 0.0%#315 | 1.1%#223 | 0.0%#315 | 5.7%#148 | – | Artificial Analysis | |
| Chess Puzzlesnot in index | 10.0%#119 | 26.0%#76 | 32.0%#64 | 36.0%#49 | – | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in index | 14.0%#97 | 16.0%#84 | 15.0%#91 | 23.0%#51 | – | Epoch AI Benchmarking Hub | |
| LMCAnot in index | – | – | – | 40.0%#107 | – | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | – | – | 90.7%#79 | – | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | – | – | 65.3%#14 | ForecastBench |
Coding
| Benchmark | minimal | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| SWE-bench Verified (Epoch) | – | 71.5%#23 | 73.5%#19 | – | Epoch AI Benchmarking Hub | |
| SciCodeeffort not stated | – | – | – | 42.9%#147 | SciCode | |
| WeirdML | – | – | 60.7%#48 | 39.8%#104 | WeirdML | |
| GSO-Bench | – | – | 6.9%#23 | – | GSO-Bench | |
| Arena Coding | – | – | 1471#117 | 1461#126 | Arena (LMArena) | |
| Arena WebDev | – | 1418#83 | – | – | Arena (LMArena) | |
| LiveCodeBench | – | – | 85.9%#27 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | – | 69.0%#64 | – | Vals AI | |
| SWE-Bench Pro | – | – | 41.8%#10 | – | Scale AI SEAL | |
| Aider Polygloteffort not stated | – | – | – | 88.0%#1 | Aider polyglot leaderboard | |
| SWE-bench Verified (bash only) | – | 65.0%#19 | – | – | SWE-bench | |
| SWE-bench Verified (any scaffold)not in indexeffort not stated | – | – | – | 74.4%#8 | SWE-bench | |
| Vibe Code Bench v1.1not in index | – | – | 20.1%#83 | – | Vals AI | |
| SWE-Bench Pro (private)not in index | – | – | 41.8%#10 | – | Scale AI SEAL | |
| ALE-Benchnot in index | 807.6#71 | – | 1162.5#46 | – | Epoch AI Benchmarking Hub | |
| AlgoTunenot in index | – | – | 1.7#8 | – | AlgoTune | |
| Terminal-Bench 2.1 (AA)not in index | – | – | 35.2%#148 | – | Artificial Analysis |
Agents & tools
| Benchmark | minimal | low | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| Terminal-Bench | – | – | 49.6%#35 | – | 49.6%#34 | Terminal-Bench | |
| GDPval | – | – | 34.8%#6 | – | – | OpenAI | |
| Remote Labor Indexeffort not stated | – | – | – | – | 1.7%#14 | Scale AI / CAIS | |
| Terminal-Bench Hard | 18.2%#171 | 26.5%#129 | 37.9%#62 | 32.6%#97 | 12.9%#206 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 67.0%#164 | 84.2%#104 | 86.5%#86 | 84.8%#97 | 0.0%#393 | Artificial Analysis | |
| τ³-Banking (AA) | – | – | – | 22.1%#89 | – | Artificial Analysis | |
| Poker Agentnot in indexeffort not stated | – | – | – | – | 1103.2%#2 | Vals AI | |
| Terminal-Bench 2.0 (Vals)not in index | – | – | – | 37.1%#38 | – | Vals AI | |
| BALROGnot in index | 32.8%#14 | – | – | – | – | BALROG | |
| DeepResearch Benchnot in index | 48.9%#16 | 49.6%#12 | 48.6%#18 | 48.1%#20 | – | DeepResearch Bench | |
| METR Time Horizonsnot in index | – | – | 69.6%#10 | 69.4%#11 | – | METR | |
| BrowseComp-Plusnot in indexeffort not stated | – | – | – | – | 57.6%#1 | BrowseComp-Plus | |
| BrowseCompnot in indexeffort not stated | – | – | – | – | 20.1%#2 | OpenAI | |
| DeepSearchQAnot in indexeffort not stated | – | – | – | – | 59.4%#5 |
Maths
| Benchmark | minimal | low | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 18.3%#95 | 37.2%#64 | – | 55.4%#49 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | 21.9%#52 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 46.7%#187 | – | 87.2%#84 | 91.4%#68 | – | Epoch AI Benchmarking Hub | |
| MATH Level 5 | – | – | 97.9%#2 | 98.1%#1 | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | 18.0%#53 | – | Vals AI | |
| AIME (Vals) | – | – | – | 93.4%#14 | – | Vals AI | |
| MGSM | – | – | – | 92.8%#14 | – | Vals AI | |
| IMO 2025possible contamination | – | – | – | 38.1%#1 | – | MathArena | |
| MathArena Apex | – | – | – | 1.0%#35 | – | MathArena | |
| Omni-MATH (HELM)effort not stated | – | – | – | – | 64.7%#7 | HELM Capabilities | |
| MATH 500not in index | – | – | – | 96.0%#3 | – | Vals AI | |
| Arena Maths | – | – | – | 1435#98 | 1410#133 | Arena (LMArena) |
Knowledge
| Benchmark | minimal | low | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | – | 50.1%#27 | – | Epoch AI Benchmarking Hub | |
| AA-Omniscience | -33.8#296 | -10.8#200 | -10.9#202 | -8.7#181 | – | Artificial Analysis | |
| MMLU-Pro | – | – | – | 86.5%#40 | – | Vals AI | |
| LegalBench | – | – | – | 86.0%#18 | – | Vals AI | |
| CorpFin | – | – | – | 61.1%#59 | – | Vals AI | |
| TaxEval | – | – | – | 73.4%#47 | – | Vals AI | |
| MedQA | – | – | – | 96.3%#4 | – | Vals AI | |
| PRBench Financeeffort not stated | – | – | – | – | 51.3%#7 | Scale AI SEAL | |
| PRBench Legaleffort not stated | – | – | – | – | 49.0%#14 | Scale AI SEAL | |
| MultiNRCeffort not statedpossible contamination | – | – | – | – | 52.1%#9 | Scale AI SEAL | |
| MMLU-Pro (HELM)not in indexeffort not stated | – | – | – | – | 86.3%#4 | HELM Capabilities | |
| CaseLaw v2not in index | – | – | – | 66.5%#6 | – | Vals AI | |
| MedCodenot in index | – | – | – | 49.6%#21 | – | Vals AI | |
| MedScribenot in index | – | – | – | 83.7%#43 | – | Vals AI | |
| EBR-benchnot in index | – | – | – | 12.7%#20 | – | Epoch AI Benchmarking Hub | |
| AA-Omniscience: accuracynot in index | 29.6%#186 | 38.1%#137 | 39.5%#128 | 40.3%#123 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 9.8%#424 | 21.0%#273 | 16.8%#321 | 17.8%#307 | – | Artificial Analysis |
Instruction following
| Benchmark | minimal | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| IFBench | 45.6%#202 | 66.6%#93 | 70.6%#66 | 73.1%#44 | – | 45.0%#208 | Artificial Analysis | |
| MultiChallenge | – | – | – | – | 63.2%#11 | – | Scale AI SEAL | |
| TutorBencheffort not stated | – | – | – | – | – | 55.3%#4 | Scale AI SEAL | |
| IFEval (HELM)effort not stated | – | – | – | – | – | 87.5%#15 | HELM Capabilities | |
| WildBench (HELM)effort not stated | – | – | – | – | – | 85.7%#6 | HELM Capabilities | |
| Arena Instruction Followingnot in index | – | – | – | 1410#123 | – | 1415#115 | Arena (LMArena) |
Human preference
| Benchmark | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|
| Arena Text | – | 1435#103 | 1426#113 | Arena (LMArena) | |
| Arena Creative Writingnot in index | – | 1373#143 | 1389#123 | Arena (LMArena) | |
| Arena Multi-turnnot in index | – | 1421#124 | 1449#86 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | – | 1413#141 | 1435#113 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 86.0%#1 | – | – | Lech Mazur |
Multimodal
| Benchmark | minimal | low | medium | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Arena Vision | – | – | – | 1208#87 | – | 1232#76 | Arena (LMArena) | |
| MMMU-Pro | 62.1%#199 | 73.8%#121 | 74.3%#112 | 74.2%#114 | – | – | Artificial Analysis | |
| VISTAeffort not stated | – | – | – | – | – | 49.7%#11 | Scale AI SEAL | |
| MMMU-Pro (Vals)not in index | – | – | – | 81.5%#39 | – | – | Vals AI | |
| MortgageTaxnot in index | – | – | – | 65.5%#41 | – | – | Vals AI | |
| SAGEnot in index | – | – | – | 43.7%#48 | – | – | Vals AI | |
| VTBnot in index | – | – | – | – | 18.7%#6 | 17.0%#7 | Scale AI SEAL | |
| GeoBenchnot in index | – | – | 81.0%#4 | – | – | – | GeoBench | |
| VPCTnot in index | – | – | 63.2%#6 | 66.0%#5 | – | – | Epoch AI Benchmarking Hub |
Long context
| Benchmark | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|
| Fiction.LiveBench 120k | 96.9%#2 | – | – | Fiction.live | |
| AA-LCR | 76.0%#148 | 78.2%#120 | 65.0%#252 | Artificial Analysis |
Composite
| Benchmark | minimal | low | medium | high | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | – | – | 150#60 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 11.4#352 | 20.8#222 | 22.9#194 | 23.0#191 | 10.4#376 | Artificial Analysis | |
| HELM Capabilities meannot in indexeffort not stated | – | – | – | – | 80.7%#4 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 17.0%#46 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 79.3%#25 | 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 | 85 tok/s | 6.11 s | $1.25 | $10.00 | 400k | – |
| Azure | 62 tok/s | 9.79 s | $1.25 | $10.00 | 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 $0.125 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0035 | $0.0032 | 1.2 min |
| Summarize a 30-page report | 12,000 / 600 | $0.021 | $0.011 | 1.3 min |
| Code edit | 6,000 / 1,500 | $0.022 | $0.017 | 1.4 min |
| Agentic coding session | 60,000 / 4,000 | $0.115 | $0.064 | 1.9 min |
| Structured extraction | 2,000 / 200 | $0.0045 | $0.0028 | 1.2 min |
More from OpenAI
See also
Questions people ask
- What are GPT-5's benchmark scores?
- As of 24 Oct 2026, GPT-5 (high effort) scores 57.9 ±2.7 on the BenchLeader Index, strongest in instruction following (72) and weakest in reasoning (53). That rests on 38 independent benchmark results, and every score on this page links to its publisher.
- How does GPT-5 rank?
- GPT-5 (high effort) is #165 of 785 ranked configurations on 24 Oct 2026. 231 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 cost per 1M tokens?
- As of 24 Oct 2026, GPT-5 costs $1.25 per 1M input tokens and $10.00 per 1M output tokens, $3.44 blended (three input tokens per output token).
- When was GPT-5 released?
- OpenAI released GPT-5 on 7 Aug 2025. BenchLeader has results for it at 6 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GPT-5: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/gpt-5.