| Model | Verdict | Decode tok/s | Max ctx | Size | Chat | Coding | |
|---|---|---|---|---|---|---|---|
| Qwen/Qwen3-30B-A3B-Instruct-2507 unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF · IQ4_XS | Fits | 36.714.7–40.4 | 36K | 30.5B | p85 | – | |
| google/gemma-4-31B-it unsloth/gemma-4-31B-it-GGUF · Q3_K_M | Fits | 12.110.3–14.0 | 53K | 31.3B | p63 | – | |
| Qwen/Qwen3.5-27B unsloth/Qwen3.5-27B-GGUF · Q4_K_M | Fits | 12.65.00–13.9 | 50K | 27.8B | p60 | p46 | |
| Qwen/Qwen3.8-27B lmstudio-community/Qwen3.8-27B-MLX-4bit · MLX-4BIT | Fits | 13.15.20–14.4 | 39K | 27.8B | p59 | p33 | |
| Qwen/Qwen3.6-27B unsloth/Qwen3.6-27B-GGUF · Q4_K_M | Fits | 12.65.00–13.8 | 48K | 27.8B | p51 | p54 | |
| Qwen/Qwen3.5-35B-A3B bartowski/Qwen_Qwen3.5-35B-A3B-GGUF · IQ3_M | Fits | 66.926.8–73.6 | 40K | 36B | p49 | p35 | |
| google/gemma-4-26B-A4B-it unsloth/gemma-4-26B-A4B-it-GGUF · MXFP4_MOE | Fits | 51.320.5–56.4 | 48K | 25.8B | p49 | – | |
| Qwen/Qwen3.5-9B lmstudio-community/Qwen3.5-9B-MLX-8bit · MLX-8BIT | Fits | 21.58.60–23.6 | 243K | 9.7B | p46 | – | |
| ornith-ai/Ornith-1.5-9B AtomicChat/Ornith-1.5-9B-GGUF · Q8_0 | Fits | 23.19.20–25.4 | 256K | 9.7B | p46 | p24 | |
| Qwen/Qwen3.6-35B-A3B unsloth/Qwen3.6-35B-A3B-GGUF · UD-IQ4_XS | Fits | 67.026.8–73.7 | 44K | 36B | p43 | p41 | |
| openbmb/MiniCPM-V-4.7-35B-A3B openbmb/MiniCPM-V-4.7-35B-A3B-gguf · IQ3_M | Fits | 70.428.1–77.4 | 128K | 35.2B | p42 | – | |
| ornith-ai/Ornith-1.0-35B bartowski/deepreinforce-ai_Ornith-1.0-35B-GGUF · Q3_K_XL | Fits | 67.827.1–74.6 | 43K | 35.1B | – | p40 | |
| openai/gpt-oss-20b unsloth/gpt-oss-20b-GGUF · F16 | Fits | 57.122.8–62.8 | 128K | 21.1B | p39 | p15 | |
| Qwen/Qwen3.5-4B unsloth/Qwen3.5-4B-GGUF · Q8_0 | Fits | 37.515.0–41.2 | 256K | 4.7B | p39 | – | |
| google/gemma-4-12B-it unsloth/gemma-4-12b-it-GGUF · Q8_0 | Fits | 13.911.8–16.0 | 256K | 12B | p38 | – | |
| Qwen/Qwen3-4B-Instruct-2507 BF16 | Fits | 19.616.7–22.6 | 71K | 4B | p38 | – | |
| nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 mlx-community/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-4bit · MLX-4BIT | Fits | 69.627.8–76.6 | 72K | 32.9B | p37 | p9 | |
| nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 unsloth/Nemotron-3-Nano-30B-A3B-GGUF · Q3_K_S | Fits | 69.027.6–76.0 | 41K | 31.6B | p33 | – | |
| poolside/Laguna-XS-2.1 bartowski/Laguna-XS-2.1-GGUF · Q3_K_XL | Fits | 74.829.9–82.3 | 46K | 33.4B | – | p30 | |
| poolside/Laguna-XS.2 manjunathshiva/Laguna-XS.2-tq3-g64 · MLX-3BIT | Fits | 77.030.8–84.7 | 80K | 33.4B | – | p25 | |
| ornith-ai/Ornith-1.0-9B lmstudio-community/Ornith-1.0-9B-MLX-8bit · MLX-8BIT | Fits | 21.58.60–23.7 | 243K | 9.4B | – | p20 | |
| Qwen/Qwen3-Coder-30B-A3B-Instruct unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF · IQ4_XS | Fits | 36.714.7–40.4 | 36K | 30.5B | – | – | |
| Qwen/Qwen3-30B-A3B unsloth/Qwen3-30B-A3B-GGUF · IQ4_XS | Fits | 36.714.7–40.4 | 36K | 30.5B | – | – | |
| prism-ml/Bonsai-27B-mlx-1bit MLX-1BIT | Fits | 29.711.9–32.7 | 199K | 26.4B | – | – | |
| ornith-ai/Ornith-1.5-9B-GGUF Q8_0 | Fits | 22.79.10–25.0 | 256K | 9.2B | – | – | |
| prism-ml/Ternary-Bonsai-27B-mlx-2bit MLX-2BIT | Fits | 21.48.50–23.5 | 150K | 26.4B | – | – | |
| Qwen/Qwen2.5-14B-Instruct Qwen/Qwen2.5-14B-Instruct-GGUF · Q6_K | Fits | 14.112.0–16.3 | 32K | 14.8B | – | – | |
| Qwen/Qwen2.5-Coder-14B-Instruct bartowski/Qwen2.5-Coder-14B-Instruct-GGUF · Q6_K_L | Fits | 13.811.7–15.8 | 32K | 14.8B | – | – | |
| Qwen/Qwen3-14B unsloth/Qwen3-14B-GGUF · Q8_0 | Fits | 12.210.4–14.1 | 35K | 14.8B | – | – | |
| farbodtavakkoli/OTel-2.0-LLM-31B-IT wfakhri/OTel-2.0-LLM-31B-IT-GGUF · Q3_K_M | Fits | 11.810.0–13.6 | 43K | 31.3B | – | – | |
| Qwen/Qwen3-32B unsloth/Qwen3-32B-GGUF · Q2_K_L | Fits | 11.710.0–13.5 | 40K | 32.8B | – | – | |
| Qwen/Qwen2.5-32B-Instruct bartowski/Qwen2.5-32B-Instruct-GGUF · IQ3_XS | Fits | 11.710.0–13.5 | 32K | 32.8B | – | – | |
| Qwen/Qwen2.5-Coder-32B-Instruct bartowski/Qwen2.5-Coder-32B-Instruct-GGUF · IQ3_XS | Fits | 11.710.0–13.5 | 32K | 32.8B | – | – |
Quality = benchmark percentile against all indexed models (p50 = median). Speeds are single-user decode ranges.