List your company

Science Question Answering on ResearchQA (accuracy)

85.8Accuracy (ResearchQA)GPT-OSS-120B
16.74434.67252.670.528May 22, 2026May 26, 2026May 30, 2026Jun 3, 2026Jun 7, 2026Jun 11, 2026Jun 15, 2026

Evaluation results

MethodLinks
GPT-OSS-120BBase Model=GPT-OSS-120B2026.0585.8
Kimi-K2.5Base Model=Kimi-K2.52026.0584.9
Deepseek-V3.2Base Model=Deepseek-V3.22026.0581.5
GLM-5.1Base Model=GLM-5.12026.0580.5
Qwen3.5-397BBase Model=Qwen3.5-397B2026.0579.4
OpenAI Deep ResearchBackbone=o3/o4, Sample size=100-sample subset2026.0679.2
Qwen3.5-122BBase Model=Qwen3.5-122B2026.0579
Qwen3.5-35BBase Model=Qwen3.5-35B2026.0578.3
GPT-5 + SearchBackbone=GPT-5, Sample size=100-sample subset2026.0678.2
Qwen3.5-9B + MaRBase Model=Qwen3.5-9B, Optimization Method=MaR2026.0576
Perplexity Deep ResearchBackbone=Proprietary2026.0675.3
DEEPRUBRIC (140-step, ≈750h)Backbone=Qwen3-8B, Steps=140-step, Compute Cost=≈750h2026.0675.2
Ai2 ScholarQABackbone=Claude Sonnet, Sample size=100-sample subset2026.0675
Qwen3.5-9BBase Model=Qwen3.5-9B2026.0574.6
Gemini 3 Pro + SearchBackbone=Gemini 3 Pro2026.0674.3
DR Tulu-8B (1900-step, ≈9700h)Backbone=Qwen3-8B, Steps=1900-step, Compute Cost=≈9700h2026.0674.3
WebThinker-32B-DPO (report)Backbone=Qwen2.5-32B (report)2026.0674.2
DEEPRUBRIC (75-step, ≈402h)Backbone=Qwen3-8B, Steps=75-step, Compute Cost=≈402h2026.0674.2
Qwen3.5-9B + DAPOBase Model=Qwen3.5-9B, Optimization Method=DAPO2026.0573.4
WebThinker (report)Backbone=QwQ-32B (report)2026.0672.8
Qwen3.5-4B + MaRBase Model=Qwen3.5-4B, Optimization Method=MaR2026.0572.7
Qwen3.5-4BBase Model=Qwen3.5-4B2026.0572.2
DR Tulu-8B (1000-step, ≈6000h)Backbone=Qwen3-8B, Steps=1000-step, Compute Cost=≈6000h2026.0671.1
Qwen3.5-4B + DAPOBase Model=Qwen3.5-4B, Optimization Method=DAPO2026.0570.4
Perplexity-Sonar (High)Backbone=Proprietary2026.0669.1
Gemini Deep ResearchBackbone=Proprietary2026.0668.5
DR Tulu-8B (SFT)Backbone=Qwen3-8B2026.0668.5
Tongyi DeepResearch-30B-A3BBackbone=Qwen2.5-30B2026.0666.7
WebExplorer-8BBackbone=Qwen2.5-8B2026.0664.8
DEEPRUBRIC (SFT)Backbone=Qwen3-8B, Stage=SFT2026.0664.8
Claude-Sonnet SearchBackbone=Claude Sonnet2026.0664.3
QwQ-32B + RAGBackbone=QwQ-32B2026.0660.9
Qwen3-8B + RAGBackbone=Qwen3-8B2026.0656.1
WebThinker-32B-DPOBackbone=Qwen2.5-32B2026.0648.6
Qwen3-8B + SearchBackbone=Qwen3-8B2026.0646.3
Search-R1-7BBackbone=Qwen2.5-7B2026.0627.9
ASearcher-Web-7BBackbone=Qwen2.5-7B2026.0619.4