Share your thoughts, 1 month free Claude Pro on usSee more

Web Task Reasoning on BrowseComp (test)

48.7Pass@1

GPT-5.4-xhigh

Updated 12d ago

Evaluation Results

Method	Links
GPT-5.4-xhigh 2026.05		48.7
MiroThinker-v1.5 2026.05		38.6
DeepSeek-V3.2 2026.05		38.1
Tongyi-DeepResearch 2026.05		36.9
Kimi-K2.5 2026.05		35.3
GLM-4.6 2026.05		22.8
SR2AM-v1.0-30B 2026.05		22.5
GPT-5.4-xhigh 2026.05		14.8
GPT-OSS-120B-high 2026.05		14.8
WebExplorer-8B 2026.05		9.4
A2FM 2026.05		7.6
ASearcher-Web-QWQ-v2 2026.05		7.1
WebSailor-32B 2026.05		6.9
DeepSeek-V3.2 2026.05		6.4
SR2AM-v0.1-8B 2026.05		4.8
WebSailor-7B 2026.05		4.7
Qwen3-235B 2026.05		4.3
K2-Think-V2-high 2026.05		2
Qwen3-30B-A3B 2026.05		1.7
ASearcher-Web-7B 2026.05		1.6
AFM-Web-7B-RL 2026.05		1.2
Qwen3-30B-A3B 2026.05		0.5
SimpleTIR-7B 2026.05		0.5
SimpleTIR-32B 2026.05		0.5
AFM-Code-7B-RL 2026.05		0