Skip to content

Benchmark

BALROG TextWorld

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
1.17

Score distribution

27.5%
Mean
18.2%
Median
23.2%
Std dev
3.3%
10th percentile
60.2%
90th percentile
0%
Min
0%75.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.17 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
1Gemini 3.1 Pro Preview· Epoch AI Benchmarking Hub data75.7%
2Claude Opus 5max· Epoch AI Benchmarking Hub data71.6%
3GPT-5.6 Solmax· Epoch AI Benchmarking Hub data63.5%
4Grok 4· Epoch AI Benchmarking Hub data62.9%
5Gemini 3 Pro Preview· Epoch AI Benchmarking Hub data60.2%
6Claude Opus 4.5· Epoch AI Benchmarking Hub data59%
7Claude Opus 4.564k· Epoch AI Benchmarking Hub data59%
8GPT-6 Astramax· Epoch AI Benchmarking Hub data54.5%
9Gemini 3 Flash· Epoch AI Benchmarking Hub data50.2%
10Gemini 2.5 Pro· Epoch AI Benchmarking Hub data49.2%
11Claude Haiku 4.5· Epoch AI Benchmarking Hub data49%
12Claude 3.5 Sonnet· Epoch AI Benchmarking Hub data42.1%
13GPT 4o· Epoch AI Benchmarking Hub data39.3%
14Grok 3beta· Epoch AI Benchmarking Hub data32.5%
15GPT 5minimal· Epoch AI Benchmarking Hub data30.6%
16Gemini 2.5 Flash· Epoch AI Benchmarking Hub data29.6%
17GPT-5.6 Terramax· Epoch AI Benchmarking Hub data27.8%
18DeepSeek R1· Epoch AI Benchmarking Hub data21.8%
19GPT-5.6 Lunamax· Epoch AI Benchmarking Hub data18.2%
20Claude Haiku 3.5· Epoch AI Benchmarking Hub data18%
21Qwen2-VL (72B) Instruct· Epoch AI Benchmarking Hub data16.5%
22Llama 3.1 70B Instruct· Epoch AI Benchmarking Hub data15%
23GPT-4o mini· Epoch AI Benchmarking Hub data12.2%
24Reka Flash 3· Epoch AI Benchmarking Hub data11.6%
25Llama 3.2 90B Vision Instruct· Epoch AI Benchmarking Hub data11.2%
26Qwen2.5 72B Instruct· Epoch AI Benchmarking Hub data11.2%
27DeepSeek R1 Distill Qwen 32B· Epoch AI Benchmarking Hub data10.6%
28Llama 3.3 70B Instruct· Epoch AI Benchmarking Hub data9%
29Llama 3.2 11B Vision Instruct· Epoch AI Benchmarking Hub data6.7%
30Llama 3.1 8B Instruct· Epoch AI Benchmarking Hub data6.1%
31Mistral Nemo· Epoch AI Benchmarking Hub data4.5%
32Qwen2.5 7B Instruct· Epoch AI Benchmarking Hub data3.9%
33Llama 3.2 3B· Epoch AI Benchmarking Hub data3.5%
34Llama 3.2 1B Instruct· Epoch AI Benchmarking Hub data3.3%
35Phi 4· Epoch AI Benchmarking Hub data2.5%
36Qwen2 VL 7b Instruct· Epoch AI Benchmarking Hub data1.6%
37gemini-1.5-flash-002· Epoch AI Benchmarking Hub data0%
38gemini-1.5-pro-002· Epoch AI Benchmarking Hub data0%

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