Skip to content

Benchmark

Arena Agent Bash Recovery Steps

Agentic benchmarks measure how well models carry out tasks that take several steps. Higher scores are better.

Models measured
50
Scores recorded
52
Sources
1
IQ weight
2.12

Score distribution

0.008
Mean
0.023
Median
0.091
Std dev
-0.075
10th percentile
0.096
90th percentile
-0.256
Min
-0.2560.15
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.12 for chat models, measured over 50 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Sonnet 5.5Max· LMArena data0.15
2Claude Opus 5Max· LMArena data0.136
3Claude Opus 5.5High· LMArena data0.136
4Claude Opus 5High· LMArena data0.126
5Claude Fable 5.1Max· LMArena data0.116
6Claude Fable 5High· LMArena data0.096
7GPT 5.5xHigh· LMArena data0.0955
8DeepSeek V4.1 FlashMax· LMArena data0.0786
9Claude Sonnet 5High· LMArena data0.0738
10Claude Opus 4.8High· LMArena data0.0722
11Grok 4.7xHigh· LMArena data0.072
12GPT-6 SolMax· LMArena data0.0703
13Hy4 preview· LMArena data0.0651
14Muse Spark 1.3Max· LMArena data0.0608
15Muse Spark 1.2xHigh· LMArena data0.0579
16Grok 4.6xHigh· LMArena data0.0568
17GPT-6 AstraMax· LMArena data0.0553
18Inkling Small· LMArena data0.0543
19GPT-6 LunaMax· LMArena data0.0521
20Grok 4.5· LMArena data0.0514
21GPT-6.1 SolMax· LMArena data0.0453
22DeepSeek V4 ProHigh) (0813· LMArena data0.044
23Kimi K3Max· LMArena data0.0438
24GPT 5.5· LMArena data0.0412
25Qwen3.8 Max· LMArena data0.028
26MiMo V2.6 Pro· LMArena data0.0234
27Mistral Large 4· LMArena data0.0178
28GPT 5.4High· LMArena data0.0121
29GPT-5.6 TerraxHigh· LMArena data0.00884
30GPT-5.6 SolxHigh· LMArena data0.00726
31GLM 5.2Max· LMArena data0.00675
32MiniMax M3· LMArena data0.0058
33GPT-5.6 LunaxHigh· LMArena data0.00514
34Step 5 Preview· LMArena data0.00228
35Inkling FP4· LMArena data0.000814
36GLM 5.3Max· LMArena data-0.000535
37Gemini 3.8 FlashHigh· LMArena data-0.00687
38MiMo V2.6 Flash· LMArena data-0.0106
39Qwen3.7 Max· LMArena data-0.0156
40Muse Spark 1.1· LMArena data-0.0226
41Qwen 3.8 Flash Next· LMArena data-0.0257
42Gemini 3.7 FlashHigh· LMArena data-0.0544
43GLM 5.3 Flash· LMArena data-0.057
44Qwen 3.8 27B· LMArena data-0.059
45Qwen3.7 Plus· LMArena data-0.0631
46Hy3 (Deep Infra)· LMArena data-0.0721
47MiMo V2.5 Pro· LMArena data-0.0751
48Gemini 3.6 FlashHigh· LMArena data-0.182
49Mistral Medium 3.5· LMArena data-0.204
50Gemini 4 ArgonHigh· LMArena data-0.214

Showing 1–50 of 52 results.

Data credits: LMArena, CC BY 4.0 (licence)