Skip to content

Benchmark

WebWalkerQA

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

Models measured
12
Scores recorded
37
Sources
1
IQ weight
0.30

Score distribution

38.7%
Mean
36%
Median
16.9%
Std dev
20%
10th percentile
61.5%
90th percentile
11.9%
Min
11.9%72.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.30 for chat models, measured over 12 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Qwen3 30B A3BTongyi DeepResearch· WebWalkerQA data72.2%
2Qwen3 30B A3BAgentFounder· WebWalkerQA data71.9%
3DeepSeek V3.1Youtu-agent· WebWalkerQA data71.5%
4Qwen2.5 32B InstructAFM-RL· WebWalkerQA data63%
5Qwen2.5 32B InstructAFM-SFT· WebWalkerQA data61.5%
6DeepSeek V3Youtu-agent· WebWalkerQA data60.7%
7Qwen2.5 7B InstructAFM-RL· WebWalkerQA data55.6%
8Qwen2.5 72B InstructWebShaper· WebWalkerQA data52.2%
9Qwen2.5 32B InstructWebShaper· WebWalkerQA data51.4%
10QwQ 32BWebShaper· WebWalkerQA data49.7%
11Qwen3 32BMiroThinker-DPO-v0.1· WebWalkerQA data49.3%
12QwQ 32BWebDancer· WebWalkerQA data47.9%
13QwQ 32BWebThinker-RL· WebWalkerQA data46.5%
14Qwen3 32BMiroThinker-SFT-v0.1· WebWalkerQA data45.7%
15Qwen3 8BMiroThinker-DPO-v0.1· WebWalkerQA data45.7%
16QwQ 32BWebThinker-Base· WebWalkerQA data41.9%
17Qwen3 8BMiroThinker-SFT-v0.1· WebWalkerQA data41.3%
18Qwen2.5 32B InstructWebDancer· WebWalkerQA data38.4%
19Qwen2.5 7B InstructWebDancer· WebWalkerQA data36%
20GPT 4oseeker· WebWalkerQA data32.2%
21GPT 4oreflexion· WebWalkerQA data30.3%
22Qwen2.5 72B Instructseeker· WebWalkerQA data29.1%
23Qwen Plusseeker· WebWalkerQA data29%
24GPT 4oreact· WebWalkerQA data28.7%
25Qwen Plusreact· WebWalkerQA data28.5%
26Qwen Plusreflexion· WebWalkerQA data28.5%
27Qwen2.5 72B Instructreflexion· WebWalkerQA data27.4%
28Qwen2.5 72B Instructreact· WebWalkerQA data26.5%
29Qwen 2.5 14B Instructseeker· WebWalkerQA data23.7%
30Qwen2.5 32B Instructseeker· WebWalkerQA data22.4%
31Qwen 2.5 14B Instructreflexion· WebWalkerQA data21.3%
32Qwen2.5 32B Instructreact· WebWalkerQA data21%
33Qwen2.5 32B Instructreflexion· WebWalkerQA data20%
34Qwen 2.5 14B Instructreact· WebWalkerQA data19.1%
35Qwen2.5 7B Instructreflexion· WebWalkerQA data15.7%
36Qwen2.5 7B Instructseeker· WebWalkerQA data15.7%
37Qwen2.5 7B Instructreact· WebWalkerQA data11.9%

Data credits: WebWalkerQA (Alibaba NLP), Apache-2.0 (licence)