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

Research Task on DeepResearchBench (DRB)

95.9Accuracy (%)

GPT 5.4

Updated 15d ago

Evaluation Results

Method	Links
GPT 5.4 2026.05		95.9	-
LLM-guided spec search 2026.05		92.5	16.7
Claude Opus 4.6 2026.05		90.1	-
Gemini 3.1 Pro 2026.05		85.2	-
Qwen3.5-122B 2026.05		80.5	-
Best Local 2026.05		80.5	-
Qwen3.5-35B 2026.05		79.4	-
Qwen3.5-27B 2026.05		77.6	-
Qwen3.5-9B 2026.05		75.8	-
Qwen3.5-9B 2026.05		75.8	-
Gemma4-26B 2026.05		72.1	-
LLM-guided spec search 2026.05		63.1	13.1
Nemotron-Super-120B 2026.05		63	-
Qwen3.5-4B 2026.05		50	-
Granite 4.0 H-Small 2026.05		42	-
LLM-guided spec search 2026.05		38.2	15.6
Gemma4-E4B 2026.05		22.6	-
Gemma4-E4B 2026.05		22.6	-
LLM-guided spec search 2026.05		15.6	15.6
Granite 3.3 8B 2026.05		10.5	-
Nemotron-Nano-4B 2026.05		0	-