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

Web Task Reasoning on XBench (test)

80.8Pass@1

GPT-5.4-xhigh

Updated 12d ago

Evaluation Results

Method	Links
GPT-5.4-xhigh 2026.05		80.8
SR2AM-v1.0-30B 2026.05		80
MiroThinker-v1.5 2026.05		76.5
Tongyi-DeepResearch 2026.05		76
Kimi-K2.5 2026.05		72
WebSailor-32B 2026.05		68.5
WebExplorer-8B 2026.05		64.5
WebSailor-7B 2026.05		63
DeepSeek-V3.2 2026.05		58.5
A2FM 2026.05		53.2
GLM-4.6 2026.05		50.5
SR2AM-v0.1-8B 2026.05		50.5
Qwen3-235B 2026.05		50.2
ASearcher-Web-QWQ-v2 2026.05		49.8
GPT-5.4-xhigh 2026.05		48.2
GPT-OSS-120B-high 2026.05		48
DeepSeek-V3.2 2026.05		37.8
Qwen3-30B-A3B 2026.05		36
AFM-Web-7B-RL 2026.05		21.2
ASearcher-Web-7B 2026.05		18.5
K2-Think-V2-high 2026.05		16.2
Qwen3-30B-A3B 2026.05		16
SimpleTIR-32B 2026.05		7
SimpleTIR-7B 2026.05		6.8
AFM-Code-7B-RL 2026.05		4