List your company

General AI Assistant Reasoning on GAIA-Text-103 1.0 (test)

62.1Overall AccuracyClaude-3.7-Sonnet
15.71627.75839.851.842Aug 1, 2025Sep 1, 2025Oct 2, 2025Nov 2, 2025Dec 3, 2025Jan 3, 2026Feb 3, 2026

Evaluation results

MethodLinks
Claude-3.7-SonnetModel Type=Proprietary2026.0262.176.957.733.3
WebShaperModel Size=72B2025.0860.169.263.416.6
WebSailorModel Size=72B2025.0855.4———
WebShaperModel Size=32B2025.0853.369.25016.6
WebSailorModel Size=32B2025.0853.2———
WebDancerModel Size=32B, Evaluation Protocol=Pass@1 (Alternative Config)2025.0851.561.55025
CSOBase Model=CK-Pro-8B, Post-Training Method=Critical Step Optimization2026.0249.561.548.116.7
Cognitive Kernel-ProModel Size=8B, Evaluation Protocol=Pass@32025.0849.361.544.216.7
WebThinker-RLModel Size=32B2025.0848.556.45016.7
GPT-4.1Model Type=Proprietary2026.0245.656.444.216.7
WebThinker-BaseModel Size=32B2025.0844.753.844.216.7
IPRBase Model=CK-Pro-8B, Post-Training Method=Iterative Process Refinement2026.0244.656.442.316.7
Cognitive Kernel-ProModel Size=8B, Backbone=Qwen-3-8B, Evaluation Protocol=Pass@12025.0843.756.442.38.33
Cognitive Kernel-ProModel Size=8B, Evaluation Protocol=Voting2025.0841.153.834.616.7
WebDancerModel Size=32B, Evaluation Protocol=Pass@12025.0840.746.144.28.3
ETOBase Model=CK-Pro-8B, Post-Training Method=Exploration-based Trajectory Optimization2026.0238.951.236.58.3
Step-DPOBase Model=CK-Pro-8B, Post-Training Method=Step-wise DPO2026.0238.953.334.68.3
WebSailorModel Size=7B2025.0837.9———
CK-Pro-8BMode=SFT2026.0235.946.234.68.3
RFTBase Model=CK-Pro-8B, Post-Training Method=Rejection Sampling Fine-Tuning2026.0234.951.228.88.3
WebDancerModel Size=7B, Evaluation Protocol=Pass@32025.0834———
WebDancerModel Size=7B, Evaluation Protocol=Pass@12025.08314130.70
Search-o1Model Size=32B2025.0828.233.3250
Qwen3-8BModel Type=Open-Source Baseline2026.0220.435.913.50
R1-SearcherModel Size=7B2025.0820.428.219.28.3
Search-o1Model Size=7B2025.0817.523.117.30