Reasoning
LiveBench Reasoning Leaderboard
LiveBench Reasoning measures logical, multi-step reasoning using contamination-free questions that are refreshed regularly, so models cannot have trained on the test set.
Source: livebench19 open models ranked+51 proprietaryData through Jun 2026
Open models ranked on LiveBench Reasoning
# shows rank among open models / rank overall (including proprietary).
| # | Model | Score | VRAM @ Q4 |
|---|---|---|---|
| 1 / 9 | Kimi K3 · 2779.9B | 90.7 | ~1674 GB |
| 2 / 28 | Qwen3.8 Flash Next · 180.0B | 87.4 | ~108 GB |
| 3 / 32 | DeepSeek V4.1 Flash · 763.2B | 86.7 | ~458 GB |
| 4 / 33 | DeepSeek V4 Flash 0731 · 304.2B | 86.6 | ~183 GB |
| 5 / 34 | Smaug Flash · 304.2B | 86.2 | ~609 GB |
| 6 / 35 | DeepSeek V4 Pro 0813 · 1650.5B | 85.8 | ~991 GB |
| 7 / 36 | GLM 5.3 · 753.3B | 85.8 | ~456 GB |
| 8 / 38 | DeepSeek V4 Flash Vision Exp · 304.6B | 85.4 | ~183 GB |
| 9 / 46 | Kimi K2.7 Code · 1026.9B | 82.8 | ~620 GB |
| 10 / 48 | DeepSeek V4 Pro · 1598.8B | 82.7 | ~960 GB |
| 11 / 55 | Qwen3.8 27B · 27.8B | 80.0 | ~17 GB |
| 12 / 56 | Kimi K2.6 · 1026.9B | 79.4 | ~620 GB |
| 13 / 57 | GLM 5.2 · 753.3B | 78.6 | ~456 GB |
| 14 / 58 | Inkling · 952.4B | 78.3 | ~572 GB |
| 15 / 60 | GLM 5.3 Flash · 321.3B | 77.6 | ~195 GB |
| 16 / 64 | NVIDIA Nemotron 3 Ultra 550B A55B BF16 · 560.5B | 74.7 | ~370 GB |
| 17 / 65 | MiniMax M3 · 427.0B | 74.5 | ~257 GB |
| 18 / 68 | DeepSeek V4 Flash · 290.9B | 70.6 | ~175 GB |
| 19 / 69 | Qwen3.6 27B · 27.8B | 70.3 | ~17 GB |
Score vs model size
Which models give the most quality for their size — the ones worth running locally.
- Qwen3.8 27B, 28B, score 80.0 — on the efficiency frontier (best score at its size or smaller).
- Qwen3.8 Flash Next, 180B, score 87.4 — on the efficiency frontier (best score at its size or smaller).
- Kimi K3, 2.8T, score 90.7 — on the efficiency frontier (best score at its size or smaller).
LiveBench Reasoning: frequently asked questions
- What is the best open LLM on LiveBench Reasoning?
- Kimi K3 is the top open model on LiveBench Reasoning, scoring 90.7. Among all models tested — including proprietary ones — it ranks #9. The top model overall is GPT 6 Astra Max (OpenAI) at 92.7.
- What's the best LiveBench Reasoning model you can run on a 24 GB GPU?
- Qwen3.8 27B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (~17 GB), scoring 80.0 on LiveBench Reasoning.
- Can open models match proprietary models on LiveBench Reasoning?
- Not quite on LiveBench Reasoning: the strongest proprietary model (GPT 6 Astra Max) scores 92.7, ahead of the best open model (Kimi K3) at 90.7 — but you can run the open one yourself.
Scores aggregated from livebench. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.