family qwen · 16 variants · unknown
| benchmark | category | best score |
|---|---|---|
| MMLU | knowledge | 92.7 |
| GPQA | reasoning | 71.1 |
| MATH | math | 2.67 |
| MT-Bench | chat | 2.21 |
| variant | kind | quant | bits | vRAM | dl | org | year |
|---|---|---|---|---|---|---|---|
| Qwen3-235B-A22B | original | bf16 | 16.0 | 806,401 | Qwen | 2025 | |
| Qwen3-235B-A22B-GGUF | gguf | gguf | 16.0 | 172,718 | unsloth | 2025 | |
| Qwen3-235B-A22B-FP8 | fp8 | fp8 | 8.0 | 165,335 | Qwen | 2025 | |
| Qwen3-235B-A22B-NVFP4 | nvfp4 | nvfp4 | 4.0 | 157,931 | nvidia | 2025 | |
| Qwen3-235B-A22B-GPTQ-Int4 | gptq | gptq | 8,685 | Qwen | 2025 | ||
| Qwen3-235B-A22B-128K-GGUF | gguf | gguf | 16.0 | 1,735 | unsloth | 2025 | |
| Qwen3-235B-A22B-GGUF | gguf | gguf | 3.0 | 1,582 | lmstudio-community | 2025 | |
| Qwen3-235B-A22B-GGUF | gguf | gguf | 4.0 | 1,068 | Qwen | 2025 | |
| Qwen3-235B-A22B-FP8 | fp8 | fp8 | 8.0 | 603 | nvidia | 2025 | |
| Qwen3-235B-A22B-Eagle3 | original | bf16 | 16.0 | 377 | nvidia | 2026 | |
| Qwen3-235B-A22B-MLX-4bit | mlx | mlx | 360 | Qwen | 2025 | ||
| Qwen3-235B-A22B-MLX-8bit | mlx | mlx | 155 | Qwen | 2025 | ||
| Qwen3-235B-A22B-MLX-bf16 | mlx | mlx | 124 | Qwen | 2025 | ||
| Qwen3-235B-A22B | original | bf16 | 16.0 | 116 | unsloth | 2025 | |
| Qwen3-235B-A22B-GGUF | gguf | gguf | 4.0 | 89 | ggml-org | 2025 | |
| Qwen3-235B-A22B-MLX-6bit | mlx | mlx | 71 | Qwen | 2025 |