Reasoning on GPQA Diamond (Score %)
85.5GPQA Diamond Reasoning ScoreQwen-3.5-27B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen-3.5-27BReasoning-mode=true2026.05 | 85.5 | |
| Gemma-4-31B2026.05 | 84.3 | |
| Qwen-3.5-9BReasoning-mode=true2026.05 | 81.7 | |
| DeepSeek-R12026.05 | 71.5 | |
| Qwen-3-32B2026.05 | 65.2 | |
| DeepSeek-V32026.05 | 59.1 | |
| Llama-3.3-70B2026.05 | 50.5 | |
| Soft+Backbone=olmo-3.1-32b-it-sft2026.04 | 49.1 | |
| Qwen-2.5-72B2026.05 | 49 | |
| Pair+Backbone=olmo-3.1-32b-it-sft2026.04 | 48.9 | |
| MPO+Backbone=olmo-3.1-32b-it-sft2026.04 | 48.5 | |
| DPOBackbone=olmo-3.1-32b-it-sft2026.04 | 47.8 | |
| Llama-3.1-70B2026.05 | 46.7 | |
| Mag+Backbone=olmo-3.1-32b-it-sft2026.04 | 46.6 | |
| Mistral-Small-3.2-24B2026.05 | 46.1 | |
| GPT-4o-Mini2026.05 | 40.2 | |
| MPO+Backbone=olmo-3-7b-it-sft2026.04 | 37.3 | |
| Mag+Backbone=olmo-3-7b-it-sft2026.04 | 36.9 | |
| Pair+Backbone=olmo-3-7b-it-sft2026.04 | 36.8 | |
| Soft+Backbone=olmo-3-7b-it-sft2026.04 | 35.4 | |
| Gemma-3-12B2026.05 | 34.9 | |
| BaseBackbone=olmo-3.1-32b-it-sft2026.04 | 34.1 | |
| Qwen-2.5-7B2026.05 | 33.8 | |
| DPOBackbone=olmo-3-7b-it-sft2026.04 | 33.7 | |
| RFTBackbone=olmo-3.1-32b-it-sft2026.04 | 33.2 | |
| Mixtral-8x22B2026.05 | 33.2 | |
| BaseBackbone=olmo-3-7b-it-sft2026.04 | 32.1 | |
| Llama-3.1-8B2026.05 | 30.4 | |
| RFTBackbone=olmo-3-7b-it-sft2026.04 | 27.9 | |
| Soft+Backbone=gemma-3-4b-sft2026.04 | 26.7 | |
| DPOBackbone=gemma-3-4b-sft2026.04 | 26.3 | |
| MPO+Backbone=gemma-3-4b-sft2026.04 | 25.5 | |
| Mag+Backbone=gemma-3-4b-sft2026.04 | 24.1 | |
| Pair+Backbone=gemma-3-4b-sft2026.04 | 23.6 | |
| BaseBackbone=gemma-3-4b-sft2026.04 | 22.3 | |
| RFTBackbone=gemma-3-4b-sft2026.04 | 21.4 |