Benchgen

PinchBench — Results

RankModelScore
1gemini-3-flash95.1
2claude-sonnet-4-593.6
3minimax-m2-193.4
4claude-haiku-4-590.2
5sage-a1-590
6kimi-k2-589.2
7claude-opus-4-588.9
8claude-opus-4-688.2
9gemini-3-pro87.7
10claude-sonnet-487.5
11nemotron-3-super-120b-a12b85.9
12gpt-4o85
13glm-4-5-air83.7
14nemotron-3-5-lightning-30b-a3b83.4
15lfm2-5-24b-a2b83.18
16glm-5-182.9
17deepseek-v3-281.1
18qwen3-5-397b-a17b80.7
19gemini-2-5-flash76.6
20gpt-4o-mini76.3
21claude-sonnet-4-668.6
22gpt-5-265.6
23minimax-m2-543.7
24glm-542.8
25step-3-5-flash40.9

PinchBench — Phases

1 phaseActive

Phase 1·

PinchBench

Aug 11, 2026
Current