Writing on WildChat 5,000 conversations benchmark leaderboard · SOTA2 Research

Writing on WildChat 5,000 conversations

0.392KLfwdUserLM-8b

Evaluation Results

Method	Links
UserLM-8b2026.05		0.392	0.424	0.411
GPT-5.42026.05		0.482	0.489	0.479
Gemini 3.1 Pro2026.05		0.482	0.471	0.466
GPT-5.4 mini2026.05		0.513	0.533	0.52
humanlm-opinion2026.05		0.514	0.519	0.531
GPT-5.4 nano2026.05		0.52	0.531	0.532
gpt-oss-120b2026.05		0.532	0.549	0.537
gemma-4-31B-it2026.05		0.537	0.547	0.545
Gemini 3 Flash2026.05		0.539	0.545	0.533
gpt-oss-20b2026.05		0.569	0.593	0.575
Qwen3.5-27B2026.05		0.57	0.58	0.584
Qwen3.5-9B2026.05		0.571	0.575	0.582
gemma-4-26B-A4B-it2026.05		0.577	0.585	0.587
Qwen3.5-122B-A10B2026.05		0.581	0.575	0.579
Gemini 3.1 Flash-Lite2026.05		0.583	0.586	0.582
gemma-4-E2B-it2026.05		0.586	0.589	0.599
gemma-4-E4B-it2026.05		0.589	0.591	0.597
Qwen3.5-4B2026.05		0.597	0.609	0.609
Llama-3.1-8B-Instruct2026.05		0.598	0.61	0.608
Llama-3.3-70B-Instruct2026.05		0.6	0.602	0.615
Qwen3.5-35B-A3B2026.05		0.602	0.609	0.615
Claude Haiku 4.52026.05		0.605	0.608	0.613
Qwen3.5 .8B2026.05		0.629	0.62	0.648
Qwen3.5-2B2026.05		0.63	0.642	0.66