List your company

Deep Search on GAIA

81.9AccuracyMiroThinker-v1.0-72B
6.801626.298345.79565.2917Jan 26, 2026Feb 15, 2026Mar 7, 2026Mar 28, 2026Apr 17, 2026May 7, 2026May 28, 2026

Evaluation results

MethodLinks
MiroThinker-v1.0-72BType=SA2026.0281.9—
TaSType=MA, Backbone=GPT-5 Medium-Think2026.0277.7—
GPT-5 High-ThinkType=N/A2026.0276.4—
MiroThinker-v1.0-30BType=SA2026.0273.5—
GPT-5 Medium-ThinkType=MA2026.0271.8—
Tongyi-DeepResearch + DS-MCMModel=Tongyi-DeepResearch, System Category=Open-Source, Search Strategy=+ DS-MCM2026.0171.7—
Tongyi DeepResearch (30B)Type=SA2026.0270.9—
OpenAI o3Model=OpenAI o3, System Category=Proprietary, Search Strategy=Vanilla2026.0170.9—
Qwen3-30B-A3B-thinking-SFT + SAPOContext Budget=128K2026.0570.9—
OpenAI DeepSearchModel=OpenAI DeepSearch, System Category=Proprietary, Search Strategy=Vanilla2026.0170.5—
OpenAI-o3Context Budget=-2026.0570.5—
Tongyi-DeepResearchModel=Tongyi-DeepResearch, System Category=Open-Source, Search Strategy=Vanilla2026.0170.1—
Tongyi-DeepResearch + LLM-CriticModel=Tongyi-DeepResearch, System Category=Open-Source, Search Strategy=+ LLM-Critic2026.0169.3—
MiroThinker-DeepResearch + DS-MCMModel=MiroThinker-DeepResearch, System Category=Open-Source, Search Strategy=+ DS-MCM2026.0169.3—
Tongyi DeepResearchContext Budget=128K2026.0568.9—
Claude-4-Sonnet (Thinking)Type=SA2026.0268.3—
OpenAI Deep ResearchType=N/A2026.0267.4—
MiroThinker-v1.0-8BType=SA2026.0266.4—
GPT-5 Medium-ThinkType=SA2026.0266—
MiroThinker-DeepResearchModel=MiroThinker-DeepResearch, System Category=Open-Source, Search Strategy=Vanilla2026.0163.8—
WebSailor-32BContext Budget=32K2026.0563.1—
MiroThinker-DeepResearch + LLM-CriticModel=MiroThinker-DeepResearch, System Category=Open-Source, Search Strategy=+ LLM-Critic2026.0163—
OpenAI GPT-5Model=OpenAI GPT-5, System Category=Proprietary, Search Strategy=Vanilla2026.0162.8—
Qwen3-30B-A3B-thinking-SFT + GRPOContext Budget=128K2026.0562.1—
Gemini 2.5 ProModel=Gemini 2.5 Pro, System Category=Proprietary, Search Strategy=Vanilla2026.0160.6—
Gemini-2.5-ProType=SA2026.0260.2—
Kimi K2Context Budget=-2026.0557.7—
Qwen3-30B-A3B-thinking-SFTContext Budget=128K2026.0557.3—
TaSType=MA, Backbone=Gemini-2.5-Flash2026.0252.4—
Qwen3-8B-SFT + SAPOContext Budget=64K2026.0552.4—
Qwen3-MaxType=MA2026.0252—
Qwen3-8B-SFT + GRPOContext Budget=64K2026.0551.5—
WebExplorer-8BContext Budget=64K2026.0550.5—
TaSType=MA, Backbone=Qwen3-Max2026.0249—
Qwen3-8B-SFT + ARPOContext Budget=64K2026.0547.3—
Qwen3-30B-MoE + DS-MCMModel=Qwen3-30B-MoE, System Category=Open-Source, Search Strategy=+ DS-MCM2026.0147.2—
Grok-3 DeeperSearchModel=Grok-3 DeeperSearch, System Category=Proprietary, Search Strategy=Vanilla2026.0147.1—
Web-30B-E-GRPOContext Budget=-2026.0546.7—
Qwen3-30B-MoE + LLM-CriticModel=Qwen3-30B-MoE, System Category=Open-Source, Search Strategy=+ LLM-Critic2026.0143.3—
Qwen3-30B-MoEModel=Qwen3-30B-MoE, System Category=Open-Source, Search Strategy=Vanilla2026.0140.9—
Qwen3-MaxType=SA2026.0239.8—
Qwen3-8B-SFTContext Budget=64K2026.0538.8—
Gemini-2.5-FlashType=MA2026.0238.4—
WebSailor-7BContext Budget=32K2026.0535.9—
MiroThinker-v1.0-8BContext Budget=64K2026.0531.2—
SAGEBackbone=QWEN-7B, Training data=SAGE (Ours), Search Engine=Google Search, Max queries=102026.0124—
MiroThinker-v1.5-30BContext Budget=128K2026.0523.3—
SAGEBackbone=QWEN-3B, Training data=SAGE (Ours), Search Engine=Google Search, Max queries=102026.0118.8—
Gemini-2.5-FlashType=SA2026.0216.3—
MusiqueBackbone=QWEN-7B, Training data=Musique, Search Engine=Google Search, Max queries=102026.0115.6—
RFTBackbone=Qwen3-8B2026.0414.98—
NQ + HotpotQABackbone=QWEN-7B, Training data=NQ + HotpotQA, Search Engine=Google Search, Max queries=102026.0114.6—
ProCeedRLBackbone=Qwen3-8B2026.0413.79—
MusiqueBackbone=QWEN-3B, Training data=Musique, Search Engine=Google Search, Max queries=102026.0113.5—
NQ + HotpotQABackbone=QWEN-3B, Training data=NQ + HotpotQA, Search Engine=Google Search, Max queries=102026.0112.5—
Qwen3-8B-v3-SFTBackbone=Qwen3-8B2026.0410.51—
DAPO/Search-R1Backbone=Qwen3-8B2026.0410.1—
Rewinding MoreBackbone=Qwen3-8B, Type=ablation2026.0410.1—
ReAct PromptingBackbone=Qwen3-8B2026.049.69—