Skip to content

Benchmark

Arena-Hard v2

People compare model answers and choose which they prefer. Scores are percentages. Higher is better.

Models measured
18
Scores recorded
21
Sources
2
IQ weight
1.89

Score distribution

68.1%
Mean
67.9%
Median
13.3%
Std dev
51.1%
10th percentile
82.7%
90th percentile
35%
Min
35%93.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.89 for chat models, measured over 18 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Step 3.5 FlashStep 3.5 technical report · Table 5 · PaCoRe· Reported by StepFun· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.93.1%
2MiMo V2 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.2%
3Qwen3 Next 80B A3B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim82.7%
4Qwen 3 235B A22B Thinking 2507· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim79.7%
5Qwen3 235B A22B Instruct 2507· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim79.2%
6Qwen3 VL 235B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.77.4%
7Step 3.5 FlashStep 3.5 technical report · Table 5 · vanilla· Reported by StepFun· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.74%
8Nemotron 3 Super 120B A12B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim73.9%
9Nemotron 3 Super 120B A12BNeMo Evaluator / mostly NeMo Skills · Hard Prompt· Reported by NVIDIA· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.9%
10Sarvam 105B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim71%
11Granite 4.2 30B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim67.9%
12Nemotron 3 Nano 30B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim67.7%
13Granite 4.2 8B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim65.2%
14Qwen3 VL 32B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.64.7%
15Qwen3 Next 80B A3Bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim62.3%
16Qwen3 VL 32B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.60.5%
17qwen3-vl-30b· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.58.5%
18qwen3-vl-30bthinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.56.7%
19Qwen: Qwen3 VL 8B Thinking· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it.51.1%
20Sarvam-30B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim49%
21Granite 4.2 3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim35%

Data credits: 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.