Skip to content

Benchmark

BALROG

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages.

Models measured
37
Scores recorded
38
Sources
1
IQ weight
1.02

Score distribution

30.8%
Mean
29.5%
Median
17.1%
Std dev
11.6%
10th percentile
57%
90th percentile
3.7%
Min
3.7%68.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.02 for chat models, measured over 37 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1GPT-6 Astramax· Epoch AI Benchmarking Hub data68.3%
2Claude Opus 5max· Epoch AI Benchmarking Hub data63.4%
3GPT-5.6 Solmax· Epoch AI Benchmarking Hub data60%
4Gemini 3 Pro Preview· Epoch AI Benchmarking Hub data58.1%
5Gemini 3.1 Pro Preview· Epoch AI Benchmarking Hub data57%
6GPT-5.6 Terramax· Epoch AI Benchmarking Hub data53.2%
7Gemini 3 Flash· Epoch AI Benchmarking Hub data48.1%
8GPT-5.6 Lunamax· Epoch AI Benchmarking Hub data45.6%
9Grok 4· Epoch AI Benchmarking Hub data43.6%
10Claude Opus 4.5· Epoch AI Benchmarking Hub data43.5%
11Gemini 2.5 Pro· Epoch AI Benchmarking Hub data43.3%
12Claude Opus 4.564k· Epoch AI Benchmarking Hub data43%
13DeepSeek R1· Epoch AI Benchmarking Hub data34.9%
14Gemini 2.5 Flash· Epoch AI Benchmarking Hub data33.5%
15GPT 5minimal· Epoch AI Benchmarking Hub data32.8%
16Claude 3.5 Sonnet· Epoch AI Benchmarking Hub data32.6%
17GPT 4o· Epoch AI Benchmarking Hub data32.3%
18Claude Haiku 4.5· Epoch AI Benchmarking Hub data31.2%
19Grok 3beta· Epoch AI Benchmarking Hub data29.5%
20Reka Flash 3· Epoch AI Benchmarking Hub data29.2%
21Llama 3.1 70B Instruct· Epoch AI Benchmarking Hub data27.9%
22Llama 3.2 90B Vision Instruct· Epoch AI Benchmarking Hub data27.3%
23Llama 3.3 70B Instruct· Epoch AI Benchmarking Hub data23%
24gemini-1.5-pro-002· Epoch AI Benchmarking Hub data21%
25DeepSeek R1 Distill Qwen 32B· Epoch AI Benchmarking Hub data19.5%
26Claude Haiku 3.5· Epoch AI Benchmarking Hub data19.3%
27Mistral Nemo· Epoch AI Benchmarking Hub data17.6%
28GPT-4o mini· Epoch AI Benchmarking Hub data17.4%
29Llama 3.2 11B Vision Instruct· Epoch AI Benchmarking Hub data16.8%
30Qwen2.5 72B Instruct· Epoch AI Benchmarking Hub data16.2%
31Llama 3.1 8B Instruct· Epoch AI Benchmarking Hub data15.1%
32gemini-1.5-flash-002· Epoch AI Benchmarking Hub data14.6%
33Qwen2-VL (72B) Instruct· Epoch AI Benchmarking Hub data12.8%
34Phi 4· Epoch AI Benchmarking Hub data11.6%
35Llama 3.2 3B· Epoch AI Benchmarking Hub data10.1%
36Qwen2.5 7B Instruct· Epoch AI Benchmarking Hub data7.8%
37Llama 3.2 1B Instruct· Epoch AI Benchmarking Hub data6.6%
38Qwen2 VL 7b Instruct· Epoch AI Benchmarking Hub data3.7%

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence)