Skip to content

Benchmark

HellaSwag

Scores are percentages. Higher is better.

Models and systems measured
56
Scores recorded
56
Sources
3
IQ weight
0.00

Score distribution

76.8%
Mean
81.5%
Median
14.8%
Std dev
55.5%
10th percentile
89.2%
90th percentile
33%
Min
33%95.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.00 for chat models, measured over 54 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1Claude 3 Opus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim95.4%
2GPT 4· Epoch AI Benchmarking Hub data
Values from 2 sources
95.3%
3GPT 4 32k· Epoch AI Benchmarking Hub data95.3%
4Gemini 1.5 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim93.3%
5MiMo V2.5 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89.8%
6Llama 3.1 405B· Epoch AI Benchmarking Hub data89.2%
7Claude 3 Sonnet· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim89%
8DeepSeek V3· Epoch AI Benchmarking Hub data88.9%
9Command R Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim88.6%
10Qwen2 72B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim87.6%
11Mixtral 8X7b V0.1· Epoch AI Benchmarking Hub data86.7%
12Gemini 1.5 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.5%
13Gemma 2 27B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.4%
14Claude 3 Haiku· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim85.9%
15Llama 3.1 Nemotron 70B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim85.6%
16Text Davinci 003· Epoch AI Benchmarking Hub data85.5%
17Qwen2.5 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim85.2%
18Qwen2.5 72B· Epoch AI Benchmarking Hub data84.8%
19phi-3.5-moe-instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83.8%
20Mistral Nemo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83.5%
21Qwen 2.5 Coder 32B· Epoch AI Benchmarking Hub data83%
22Qwen 2.5 Coder 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83%
23Phi-3 Medium 128K Instruct· Epoch AI Benchmarking Hub data82.4%
24gemma-7b· Epoch AI Benchmarking Hub data82.2%
25Gemma 2 9B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.9%
26Text Davinci 002· Epoch AI Benchmarking Hub data81.5%
27Mistral 7B v0.1· Epoch AI Benchmarking Hub data81%
28Llama 2 13b· Epoch AI Benchmarking Hub data80.7%
29qwen2p5-coder-14b· Epoch AI Benchmarking Hub data80.2%
30text-davinci-001· Epoch AI Benchmarking Hub data79.3%
31LLaMA 13B· Epoch AI Benchmarking Hub data79.2%
32davinci· Epoch AI Benchmarking Hub data77.5%
33Llama 2 7b· Epoch AI Benchmarking Hub data77.2%
34Phi-3 Small 8K Instruct· Epoch AI Benchmarking Hub data77%
35Qwen2.5 Coder 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim76.8%
36Qwen2.5-Coder-7B· Epoch AI Benchmarking Hub data76.8%
37Phi-3 Mini 4K Instruct· Epoch AI Benchmarking Hub data76.7%
38Yi-6B· Epoch AI Benchmarking Hub data74.4%
39gemma-2b· Epoch AI Benchmarking Hub data71.4%
40qwen2p5-coder-3b· Epoch AI Benchmarking Hub data70.9%
41Llama 3.2 3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim69.8%
42Phi 4 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim69.1%
43curie· Epoch AI Benchmarking Hub data68.2%
44Text Curie 001· Epoch AI Benchmarking Hub data67.6%
45qwen2p5-coder-1p5b· Epoch AI Benchmarking Hub data61.8%
46ChatGLM2 6B· Epoch AI Benchmarking Hub data57%
47Text Babbage 001· Epoch AI Benchmarking Hub data56.1%
48babbage· Epoch AI Benchmarking Hub data55.5%
49phi-2· Epoch AI Benchmarking Hub data53.6%
50qwen2p5-coder-0p5b· Epoch AI Benchmarking Hub data48.4%

Showing 1–50 of 54 results.

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
DeepSeek V4.1 Flash BaseEM · 0-shot· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.87.2%
MiMo V2 Flash Baseofficial Base table · 10-shot· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.88.5%

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.