physics benchmark
GPQA Leaderboard
A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy.
Updated Aug 17, 2026
Models100
Model coverage239
MetricScore
EvidenceC
GPQA Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAN | Score94.6% | Percentile100.0% | Participants239 | EvidenceC | Evaluated |
| Rank02 | ModelOP | Score94.6% | Percentile99.6% | Participants239 | EvidenceC | Evaluated |
| Rank03 | ModelGO | Score94.3% | Percentile99.2% | Participants239 | EvidenceC | Evaluated |
| Rank04 | ModelAN | Score94.2% | Percentile98.7% | Participants239 | EvidenceC | Evaluated |
| Rank05 | ModelAN | Score93.6% | Percentile98.3% | Participants239 | EvidenceC | Evaluated |
| Rank06 | ModelOP | Score93.6% | Percentile97.9% | Participants239 | EvidenceC | Evaluated |
| Rank07 | ModelMA | Score93.5% | Percentile97.5% | Participants239 | EvidenceC | Evaluated |
| Rank08 | ModelOP | Score93.2% | Percentile97.1% | Participants239 | EvidenceC | Evaluated |
| Rank09 | ModelXA | Score93.0% | Percentile96.6% | Participants239 | EvidenceB | Evaluated |
| Rank10 | ModelOP | Score92.9% | Percentile96.2% | Participants239 | EvidenceC | Evaluated |
| Rank11 | ModelOP | Score92.8% | Percentile95.8% | Participants239 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score92.6% | Percentile95.4% | Participants239 | EvidenceC | Evaluated |
| Rank13 | ModelOP | Score92.4% | Percentile95.0% | Participants239 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score92.4% | Percentile94.5% | Participants239 | EvidenceC | Evaluated |
| Rank15 | ModelOP | Score92.3% | Percentile94.1% | Participants239 | EvidenceC | Evaluated |
| Rank16 | ModelGO | Score91.9% | Percentile93.7% | Participants239 | EvidenceC | Evaluated |
| Rank17 | ModelAN | Score91.3% | Percentile93.3% | Participants239 | EvidenceC | Evaluated |
| Rank18 | ModelZA | Score91.2% | Percentile92.9% | Participants239 | EvidenceC | Evaluated |
| Rank19 | ModelMA | Score90.5% | Percentile92.4% | Participants239 | EvidenceC | Evaluated |
| Rank20 | ModelGO | Score90.4% | Percentile92.0% | Participants239 | EvidenceC | Evaluated |
| Rank21 | ModelTE | Score90.4% | Percentile91.6% | Participants239 | EvidenceC | Evaluated |
| Rank22 | ModelAC | Score90.4% | Percentile91.2% | Participants239 | EvidenceC | Evaluated |
| Rank23 | ModelAC | Score90.3% | Percentile90.8% | Participants239 | EvidenceC | Evaluated |
| Rank24 | ModelDE | Score90.1% | Percentile90.3% | Participants239 | EvidenceC | Evaluated |
| Rank25 | ModelAN | Score89.9% | Percentile89.9% | Participants239 | EvidenceC | Evaluated |
| Rank26 | ModelTM | Score89.5% | Percentile89.5% | Participants239 | EvidenceC | Evaluated |
| Rank27 | ModelME | Score89.5% | Percentile89.1% | Participants239 | EvidenceC | Evaluated |
| Rank28 | ModelAC | Score89.2% | Percentile88.7% | Participants239 | EvidenceC | Evaluated |
| Rank29 | ModelUP | Score89.0% | Percentile88.2% | Participants239 | EvidenceC | Evaluated |
| Rank30 | ModelBY | Score88.9% | Percentile87.8% | Participants239 | EvidenceC | Evaluated |
| Rank31 | ModelXA | Score88.4% | Percentile87.4% | Participants239 | EvidenceC | Evaluated |
| Rank32 | ModelAC | Score88.4% | Percentile87.0% | Participants239 | EvidenceC | Evaluated |
| Rank33 | ModelDE | Score88.1% | Percentile86.5% | Participants239 | EvidenceC | Evaluated |
| Rank34 | ModelOP | Score88.1% | Percentile86.1% | Participants239 | EvidenceC | Evaluated |
| Rank35 | ModelOP | Score88.1% | Percentile85.7% | Participants239 | EvidenceC | Evaluated |
| Rank36 | ModelOP | Score88.1% | Percentile85.3% | Participants239 | EvidenceC | Evaluated |
| Rank37 | ModelOP | Score88.1% | Percentile84.9% | Participants239 | EvidenceC | Evaluated |
| Rank38 | ModelOP | Score88.1% | Percentile84.5% | Participants239 | EvidenceC | Evaluated |
| Rank39 | ModelOP | Score88.0% | Percentile84.0% | Participants239 | EvidenceC | Evaluated |
| Rank40 | ModelAC | Score87.8% | Percentile83.6% | Participants239 | EvidenceC | Evaluated |
| Rank41 | ModelMA | Score87.6% | Percentile83.2% | Participants239 | EvidenceC | Evaluated |
| Rank42 | ModelXA | Score87.5% | Percentile82.8% | Participants239 | EvidenceC | Evaluated |
| Rank43 | ModelDE | Score87.4% | Percentile82.3% | Participants239 | EvidenceC | Evaluated |
| Rank44 | ModelOP | Score87.3% | Percentile81.9% | Participants239 | EvidenceC | Evaluated |
| Rank45 | ModelAN | Score87.0% | Percentile81.5% | Participants239 | EvidenceC | Evaluated |
| Rank46 | ModelNV | Score87.0% | Percentile81.1% | Participants239 | EvidenceC | Evaluated |
| Rank47 | ModelGO | Score86.9% | Percentile80.7% | Participants239 | EvidenceC | Evaluated |
| Rank48 | ModelAC | Score86.6% | Percentile80.3% | Participants239 | EvidenceC | Evaluated |
| Rank49 | ModelGO | Score86.4% | Percentile79.8% | Participants239 | EvidenceC | Evaluated |
| Rank50 | ModelZA | Score86.2% | Percentile79.4% | Participants239 | EvidenceC | Evaluated |
| Rank51 | ModelAC | Score86.0% | Percentile79.0% | Participants239 | EvidenceC | Evaluated |
| Rank52 | ModelZA | Score85.7% | Percentile78.6% | Participants239 | EvidenceC | Evaluated |
| Rank53 | ModelOP | Score85.7% | Percentile78.2% | Participants239 | EvidenceC | Evaluated |
| Rank54 | ModelXA | Score85.7% | Percentile77.7% | Participants239 | EvidenceC | Evaluated |
| Rank55 | ModelOP | Score85.6% | Percentile77.3% | Participants239 | EvidenceC | Evaluated |
| Rank56 | ModelAC | Score85.5% | Percentile76.9% | Participants239 | EvidenceC | Evaluated |
| Rank57 | ModelBY | Score85.1% | Percentile76.5% | Participants239 | EvidenceC | Evaluated |
| Rank58 | ModelBA | Score85.0% | Percentile76.0% | Participants239 | EvidenceC | Evaluated |
| Rank59 | ModelAN | Score84.8% | Percentile75.6% | Participants239 | EvidenceC | Evaluated |
| Rank60 | ModelXA | Score84.6% | Percentile75.2% | Participants239 | EvidenceC | Evaluated |
| Rank61 | ModelMI | Score84.6% | Percentile74.8% | Participants239 | EvidenceC | Evaluated |
| Rank62 | ModelMA | Score84.5% | Percentile74.4% | Participants239 | EvidenceC | Evaluated |
| Rank63 | ModelGO | Score84.3% | Percentile74.0% | Participants239 | EvidenceC | Evaluated |
| Rank64 | ModelMI | Score84.2% | Percentile73.5% | Participants239 | EvidenceC | Evaluated |
| Rank65 | ModelAC | Score84.2% | Percentile73.1% | Participants239 | EvidenceC | Evaluated |
| Rank66 | ModelOP | Score84.0% | Percentile72.7% | Participants239 | EvidenceC | Evaluated |
| Rank67 | ModelXA | Score84.0% | Percentile72.3% | Participants239 | EvidenceC | Evaluated |
| Rank68 | ModelXI | Score83.7% | Percentile71.8% | Participants239 | EvidenceC | Evaluated |
| Rank69 | ModelME | Score83.5% | Percentile71.4% | Participants239 | EvidenceC | Evaluated |
| Rank70 | ModelAN | Score83.4% | Percentile71.0% | Participants239 | EvidenceC | Evaluated |
| Rank71 | ModelOP | Score83.3% | Percentile70.6% | Participants239 | EvidenceC | Evaluated |
| Rank72 | ModelGO | Score83.0% | Percentile70.2% | Participants239 | EvidenceC | Evaluated |
| Rank73 | ModelGO | Score82.8% | Percentile69.8% | Participants239 | EvidenceC | Evaluated |
| Rank74 | ModelOP | Score82.8% | Percentile69.3% | Participants239 | EvidenceC | Evaluated |
| Rank75 | ModelNV | Score82.7% | Percentile68.9% | Participants239 | EvidenceC | Evaluated |
| Rank76 | ModelDE | Score82.4% | Percentile68.5% | Participants239 | EvidenceC | Evaluated |
| Rank77 | ModelDE | Score82.4% | Percentile68.1% | Participants239 | EvidenceC | Evaluated |
| Rank78 | ModelGO | Score82.3% | Percentile67.7% | Participants239 | EvidenceC | Evaluated |
| Rank79 | ModelOP | Score82.3% | Percentile67.2% | Participants239 | EvidenceC | Evaluated |
| Rank80 | ModelAC | Score81.7% | Percentile66.8% | Participants239 | EvidenceC | Evaluated |
| Rank81 | ModelME | Score81.5% | Percentile66.4% | Participants239 | EvidenceC | Evaluated |
| Rank82 | ModelAM Nova 2 ProAmazon | Score81.4% | Percentile66.0% | Participants239 | EvidenceC | Evaluated |
| Rank83 | ModelOP | Score81.4% | Percentile65.5% | Participants239 | EvidenceC | Evaluated |
| Rank84 | ModelAC | Score81.1% | Percentile65.1% | Participants239 | EvidenceC | Evaluated |
| Rank85 | ModelDE | Score81.0% | Percentile64.7% | Participants239 | EvidenceC | Evaluated |
| Rank86 | ModelZA | Score81.0% | Percentile64.3% | Participants239 | EvidenceC | Evaluated |
| Rank87 | ModelMI | Score81.0% | Percentile63.9% | Participants239 | EvidenceC | Evaluated |
| Rank88 | ModelAN | Score80.9% | Percentile63.5% | Participants239 | EvidenceC | Evaluated |
| Rank89 | ModelOP | Score80.9% | Percentile63.0% | Participants239 | EvidenceC | Evaluated |
| Rank90 | ModelME | Score80.5% | Percentile62.6% | Participants239 | EvidenceC | Evaluated |
| Rank91 | ModelOP | Score80.1% | Percentile62.2% | Participants239 | EvidenceC | Evaluated |
| Rank92 | ModelDE | Score79.9% | Percentile61.8% | Participants239 | EvidenceC | Evaluated |
| Rank93 | ModelAN | Score79.6% | Percentile61.3% | Participants239 | EvidenceC | Evaluated |
| Rank94 | ModelAMNova 2 LiteAmazon | Score79.6% | Percentile60.9% | Participants239 | EvidenceC | Evaluated |
| Rank95 | ModelZA | Score79.1% | Percentile60.5% | Participants239 | EvidenceC | Evaluated |
| Rank96 | ModelOP | Score79.0% | Percentile60.1% | Participants239 | EvidenceC | Evaluated |
| Rank97 | ModelGO | Score78.8% | Percentile59.7% | Participants239 | EvidenceC | Evaluated |
| Rank98 | ModelSA | Score78.7% | Percentile59.2% | Participants239 | EvidenceC | Evaluated |
| Rank99 | ModelMI | Score78.0% | Percentile58.8% | Participants239 | EvidenceC | Evaluated |
| Rank100 | ModelOP | Score78.0% | Percentile58.4% | Participants239 | EvidenceC | Evaluated |
GPQA Highlights
The leading models and scores on this benchmark.
GPQA Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for GPQA
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis gpqa AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01ANClaude Mythos PreviewAnthropic
- Score
- 94.6%
Strengths
- Ranks #1 of 239 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures GPQA, not total model capability
- 02OPGPT-5.6 SolOpenAI
- Score
- 94.6%
- Price
- $5.0 input / $30 output per 1M tokens
- Speed
- Up to 27 tok/s via OpenAI
Strengths
- Ranks #2 of 239 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures GPQA, not total model capability
- 03GOGemini 3.1 ProGoogle
- Score
- 94.3%
- Price
- $2.0 input / $12 output per 1M tokens
- Speed
- Up to 23 tok/s via Google
Strengths
- Ranks #3 of 239 compared models
- 99th percentile on this benchmark
- C evidence result
Considerations
- This result measures GPQA, not total model capability
- 04ANClaude Opus 4.7Anthropic
- Score
- 94.2%
- Price
- $5.0 input / $25 output per 1M tokens
- Speed
- Up to 42 tok/s via Anthropic
Strengths
- Ranks #4 of 239 compared models
- 99th percentile on this benchmark
- C evidence result
Considerations
- This result measures GPQA, not total model capability
- 05ANClaude Opus 4.8Anthropic
- Score
- 93.6%
- Price
- $5.0 input / $25 output per 1M tokens
- Speed
- Up to 46 tok/s via Anthropic
Strengths
- Ranks #5 of 239 compared models
- 98th percentile on this benchmark
- C evidence result
Considerations
- This result measures GPQA, not total model capability
Selection summary
Best AI Models for GPQA
Claude Mythos Preview currently leads GPQA with 94.6%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is GPQA?
What GPQA measures and how its scores work.
A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of C.
- Family
- GPQA
- Modality
- text
- Primary category
- physics
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- gpqa|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about GPQA.
Which model scores highest on GPQA?
Claude Mythos Preview is currently ranked first with 94.6%.
What does GPQA measure?
A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
100 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.