List your company

Multi-turn task completion on TelepathyGym

90.2Task Completion ScoreGemini-1.5-Pro
1.48824.51947.5570.581Apr 8, 2026

Evaluation results

MethodLinks
Gemini-1.5-ProCategory=Closed-Source LLM2026.0490.2
GPT-4oCategory=Closed-Source LLM2026.0485.4
Gemini-1.5-FlashCategory=Closed-Source LLM2026.0463.4
GRPO + Qwen2.5-4BCategory=RL (GRPO) Trained Models2026.0463.4
Qwen2.5-14B + PRIMECategory=PRIME Models2026.0463.2
Qwen2.5-14BCategory=Open-Source LLM2026.0458.5
Qwen2.5-32BCategory=Open-Source LLM2026.0456.1
GRPO + Qwen2.5-8BCategory=RL (GRPO) Trained Models2026.0456.1
Qwen2.5-4B + PRIMECategory=PRIME Models2026.0455.8
Qwen2.5-8B + PRIMECategory=PRIME Models2026.0452.4
Qwen2.5-8BCategory=Open-Source LLM2026.0451
Qwen2.5-4BCategory=Open-Source LLM2026.0448.8
GPT-4o-miniCategory=Closed-Source LLM2026.044.9