llmrun
Reasoning

SimpleBench Leaderboard

SimpleBench is a set of everyday, common-sense and trick questions that humans answer easily but language models often get wrong. It probes basic reasoning and robustness rather than specialist knowledge.

Source: epoch23 open models ranked+78 proprietaryData through Aug 2026

Open modelsAll models

Open models ranked on SimpleBench

# shows rank among open models / rank overall (including proprietary).

#ModelScore
1 / 27DeepSeek V4 Flash 0731 · 304.2B
61.1%
2 / 29Kimi K3 · 2779.9B
60.7%
3 / 36GLM 5.2 · 753.3B
58.8%
4 / 37Kimi K2.7 Code · 1026.9B
57.9%
5 / 42GLM 5.1 · 753.9B
55.1%
6 / 45GLM 5 · 753.9B
53.2%
7 / 48DeepSeek V3.2 Speciale · 685.4B
52.6%
8 / 51Inkling · 952.4B
50.0%
9 / 54GLM 4.7 · 358.3B
47.7%
10 / 57Kimi K2.5 · 1026.9B
46.8%
11 / 60DeepSeek V4 Flash · 290.9B
46.3%
12 / 63MiniMax M3 · 427.0B
45.8%
13 / 70DeepSeek R1 0528 · 684.5B
40.8%
14 / 72DeepSeek V3.1 · 684.5B
40.0%
15 / 79Qwen3 235B A22B · 235.1B
31.0%
16 / 80DeepSeek R1 · 684.5B
30.9%
17 / 82Llama 4 Maverick 17B 128E Instruct · 401.6B
27.7%
18 / 84DeepSeek v3 0324 · 684.5B
27.2%
19 / 87Kimi K2 Instruct · 1026.4B
26.3%
20 / 90Llama 3.1 405B Instruct · 405.9B
23.0%
21 / 94GPT OSS 120B · 116.8B
22.1%
22 / 95Llama 3.3 70B Instruct · 70.6B
19.9%
23 / 96DeepSeek v3 · 684.5B
18.9%

Score vs model size

Which models give the most quality for their size — the ones worth running locally.

100B1Tmodel size (log scale) →61.1%18.9%Kimi K3 · 2.8T · 60.7%GLM 5.2 · 753B · 58.8%Kimi K2.7 Code · 1T · 57.9%GLM 5.1 · 754B · 55.1%GLM 5 · 754B · 53.2%DeepSeek V3.2 Speciale · 685B · 52.6%Inkling · 952B · 50.0%GLM 4.7 · 358B · 47.7%Kimi K2.5 · 1T · 46.8%MiniMax M3 · 427B · 45.8%DeepSeek R1 0528 · 685B · 40.8%DeepSeek V3.1 · 685B · 40.0%DeepSeek R1 · 685B · 30.9%Llama 4 Maverick 17B 128E Instruct · 402B · 27.7%DeepSeek v3 0324 · 685B · 27.2%Kimi K2 Instruct · 1T · 26.3%Llama 3.1 405B Instruct · 406B · 23.0%DeepSeek v3 · 685B · 18.9%Llama 3.3 70B Instruct · 71B · 19.9%Llama 3.3 70B InstructGPT OSS 120B · 117B · 22.1%GPT OSS 120BQwen3 235B A22B · 235B · 31.0%Qwen3 235B A22BDeepSeek V4 Flash · 291B · 46.3%DeepSeek V4 FlashDeepSeek V4 Flash 0731 · 304B · 61.1%DeepSeek V4 Flash 0731
Each dot is a model. Up = higher score, left = smaller (easier to run locally). The dashed line marks the efficiency frontier — the best score you can get at each size or smaller.
  • Llama 3.3 70B Instruct, 71B, score 19.9% — on the efficiency frontier (best score at its size or smaller).
  • GPT OSS 120B, 117B, score 22.1% — on the efficiency frontier (best score at its size or smaller).
  • Qwen3 235B A22B, 235B, score 31.0% — on the efficiency frontier (best score at its size or smaller).
  • DeepSeek V4 Flash, 291B, score 46.3% — on the efficiency frontier (best score at its size or smaller).
  • DeepSeek V4 Flash 0731, 304B, score 61.1% — on the efficiency frontier (best score at its size or smaller).

SimpleBench: frequently asked questions

What is the best open LLM on SimpleBench?
DeepSeek V4 Flash 0731 is the top open model on SimpleBench, scoring 61.1%. Among all models tested — including proprietary ones — it ranks #27. The top model overall is Claude Fable 5 Max (Anthropic) at 81.9%.
Can open models match proprietary models on SimpleBench?
Not quite on SimpleBench: the strongest proprietary model (Claude Fable 5 Max) scores 81.9%, ahead of the best open model (DeepSeek V4 Flash 0731) at 61.1% — but you can run the open one yourself.

Scores aggregated from epoch. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.