List your company

Long-horizon agentic tasks on BrowseComp Full

56.1Pass@1MiroThinker-v1.5-30B-A3B
3.16416.90730.6544.393Mar 29, 2026

Evaluation results

MethodLinks
MiroThinker-v1.5-30B-A3BContext Management=-2026.0356.1
OpenAI DeepResearchContext Management=-2026.0351.5
DeepSeek-v3.2Context Management=Baseline (w/o CM)2026.0351.4
GPT-5.1 HighContext Management=w/o CM2026.0350.8
OpenAI-o3Context Management=w/o CM2026.0349.7
Tongyi-DR-30B-A3BContext Management=Baseline (w/o CM)2026.0343.4
AgentFounder-30B-A3BContext Management=-2026.0339.9
Gemini-3.0-ProContext Management=w/o CM2026.0337.8
IterResearch-30B-A3BContext Management=-2026.0337.3
Claude-4.5-OpusContext Management=w/o CM2026.0337
AgentFold-30B-A3BContext Management=-2026.0336.2
DeepMiner-32B-RLContext Management=-2026.0333.5
ASearcher-Web-32BContext Management=-2026.035.2