Benchgen

HellaSwag — Results

RankModelScore
1claude-3-5-sonnet95.4
2claude-3-opus95.4
3gpt-4-061395.3
4gpt-4o95.3
5llama-3-1-70b-instruct93.8
6gemini-1-5-pro92.5
7llama-3-1-405b-instruct92.2
8mimo-v2-5-pro89.8
9claude-3-sonnet89
10command-r-plus88.6
11hermes-3-70b88.2
12qwen2-72b-instruct87.6
13gemini-1-5-flash86.5
14gemma-2-27b86.4
15claude-3-haiku85.9
16llama-3-1-nemotron-70b-instruct85.6
17qwen2-5-32b-instruct85.2
18phi-3-5-moe-instruct83.8
19mistral-nemo-instruct83.5
20qwen2-5-coder-32b-instruct83
21gemma-2-9b81.9
22granite-3-3-8b-base80.1
23gemma-3n-e4b-litert-preview78.6
24gemma-3n-e4b78.6
25qwen2-5-coder-7b-instruct76.8
H

HellaSwag — Phases

1 phaseActive

Phase 1·

HellaSwag

Jul 23, 2026
Current