LocalLLMLeaderboard
Home / By Size / 32b

Best 32b Models in 2026 (Updated Weekly)

As of 2026-W40, the best 32b model is DeepSeek R1 Distill Qwen 32B GGUF (Intelligence 8.4).
#ModelParamsIntelligenceCodingDownloads 30d
1DeepSeek R1 Distill Qwen 32B GGUF32.8B8.4—0.05M
2Qwen2.5 Coder 32B Instruct32.8B6.7—1.05M
3Qwen2.5 Coder 32B Instruct abliterated GGUF32.8B6.7—0.01M
4Qwen3.8 Flash Next MLX Serve mixed 4 8bit133.2B——0.01M
5llm jp 4.1 32b a3b thinking gguf32.1B——0.00M
6llm jp 4.1 32b a3b thinking32.1B——0.00M
7Qwen3 VL 32B Heretic MiniMax H3 NVFP4————
8Qwen3 VL 32B Ultra Heretic H3 ComfyUI INT8 ConvRot————
9Qwen3 VL 32B Heretic MiniMax H3 NVFP4————