# GPQA Diamond — AI model rankings

Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.

46 tracked models have a published GPQA Diamond score. Higher is better. Scores are as published at each model's release.

## Ranking

| Rank | Model | Developer | Score | Released |
| --- | --- | --- | --- | --- |
| 1 | GPT-5.4-Pro | OpenAI | 94.4% | Mar 5 2026 |
| 2 | Gemini 3.1 Pro | Google | 94.3% | Feb 19 2026 |
| 3 | Claude Opus 4.7 | Anthropic | 94.2% | Apr 16 2026 |
| 4 | GPT-5.5 | OpenAI | 93.6% | Apr 23 2026 |
| 5 | Kimi K3 | Moonshot AI | 93.5% | Jul 16 2026 |
| 6 | GPT-5.4 | OpenAI | 92.8% | Mar 5 2026 |
| 7 | GPT-5.2 | OpenAI | 92.4% | Dec 11 2025 |
| 7 | Qwen3.7-Max | Qwen | 92.4% | May 20 2026 |
| 9 | Gemini 3.0 Pro | Google | 91.9% | Nov 18 2025 |
| 10 | Claude Opus 4.6 | Anthropic | 91.3% | Feb 5 2026 |
| 11 | GLM-5.2 | Z.ai | 91.2% | Jun 16 2026 |
| 12 | Kimi K2.6 | Moonshot AI | 90.5% | Apr 21 2026 |
| 13 | Gemini 3.0 Flash | Google | 90.4% | Dec 17 2025 |
| 14 | DeepSeek-V4-Pro | DeepSeek | 90.1% | Apr 24 2026 |
| 15 | Claude Sonnet 4.6 | Anthropic | 89.9% | Feb 17 2026 |
| 16 | Muse Spark | Meta | 89.5% | Apr 8 2026 |
| 17 | Qwen3.5 | Qwen | 88.4% | Feb 16 2026 |
| 18 | GPT-5.1 | OpenAI | 88.1% | Nov 12 2025 |
| 18 | DeepSeek-V4-Flash | DeepSeek | 88.1% | Apr 24 2026 |
| 20 | Kimi K2.5 | Moonshot AI | 87.6% | Jan 27 2026 |
| 21 | Claude Opus 4.5 | Anthropic | 87% | Nov 24 2025 |
| 22 | Gemini 2.5 Pro | Google | 86.4% | Mar 25 2025 |
| 23 | GLM-5.1 | Z.ai | 86.2% | Apr 7 2026 |
| 24 | GLM-5 | Z.ai | 86% | Feb 12 2026 |
| 24 | Qwen3.6 | Qwen | 86% | Apr 16 2026 |
| 26 | GLM-4.7 | Z.ai | 85.7% | Dec 22 2025 |
| 27 | Claude Sonnet 4.5 | Anthropic | 83.4% | Sep 29 2025 |
| 28 | Claude Opus 4.1 | Anthropic | 80.9% | Aug 5 2025 |
| 29 | Claude Opus 4 | Anthropic | 79.6% | May 22 2025 |
| 30 | GLM-4.5 | Z.ai | 79.1% | Jul 28 2025 |
| 31 | o1 | OpenAI | 75.7% | Dec 5 2024 |
| 32 | Claude Sonnet 4 | Anthropic | 75.4% | May 22 2025 |
| 33 | Kimi K2 | Moonshot AI | 75.1% | Jul 11 2025 |
| 33 | Kimi K2 (0905) | Moonshot AI | 75.1% | Sep 5 2025 |
| 35 | o1-preview | OpenAI | 73.3% | Sep 12 2024 |
| 36 | Claude Haiku 4.5 | Anthropic | 73% | Oct 15 2025 |
| 37 | Claude 3.7 Sonnet | Anthropic | 68% | Feb 24 2025 |
| 38 | o1-mini | OpenAI | 60% | Sep 12 2024 |
| 39 | Claude 3.5 Sonnet | Anthropic | 59.4% | Jun 20 2024 |
| 40 | Claude 3 Opus | Anthropic | 50.4% | Mar 4 2024 |
| 41 | GPT-4o | OpenAI | 49.9% | May 13 2024 |
| 42 | GPT-4 Turbo | OpenAI | 42.5% | Nov 6 2023 |
| 43 | Claude 3.5 Haiku | Anthropic | 41.6% | Oct 22 2024 |
| 44 | Claude 3 Sonnet | Anthropic | 40.4% | Mar 4 2024 |
| 45 | GPT-4o mini | OpenAI | 40.2% | Jul 18 2024 |
| 46 | Claude 3 Haiku | Anthropic | 33.3% | Mar 4 2024 |


---

Canonical page: https://aireleasetracker.com/benchmark/gpqa
Full dataset: https://aireleasetracker.com/llms-full.txt · JSON: https://aireleasetracker.com/models.json
Source: AI Release Tracker (https://aireleasetracker.com). Benchmark scores are the figures published by the releasing lab at launch.