| Model | Verdict | Decode tok/s | Max ctx | Size | Chat | Coding | |
|---|---|---|---|---|---|---|---|
| Qwen/Qwen3-30B-A3B-Instruct-2507 unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF · Q3_K_S | Fits | 80.432.2–88.4 | 256K | 30.5B | p85 | – | |
| google/gemma-4-31B-it unsloth/gemma-4-31B-it-GGUF · Q3_K_S | Fits | 20.38.10–22.3 | 38K | 31.3B | p63 | – | |
| Qwen/Qwen3.5-27B unsloth/Qwen3.5-27B-GGUF · IQ4_XS | Fits | 22.38.90–24.6 | 38K | 27.8B | p60 | p46 | |
| Qwen/Qwen3.8-27B unsloth/Qwen3.8-27B-GGUF · UD-IQ4_XS | Fits | 23.19.30–25.5 | 53K | 27.8B | p59 | p33 | |
| Qwen/Qwen3.6-27B unsloth/Qwen3.6-27B-GGUF · Q3_K_M | Fits | 24.39.70–26.7 | 40K | 27.8B | p51 | p54 | |
| Qwen/Qwen3.5-35B-A3B unsloth/Qwen3.5-35B-A3B-GGUF · UD-IQ3_S | Fits | 13152.3–144 | 256K | 36B | p49 | p35 | |
| google/gemma-4-26B-A4B-it unsloth/gemma-4-26B-A4B-it-GGUF · UD-IQ4_NL | Fits | 55.322.1–60.8 | 256K | 25.8B | p49 | – | |
| Qwen/Qwen3.5-9B unsloth/Qwen3.5-9B-GGUF · Q8_0 | Fits | 35.814.3–39.4 | 214K | 9.7B | p46 | – | |
| ornith-ai/Ornith-1.5-9B bartowski/Ornith-1.5-9B-GGUF · Q8_0 | Fits | 35.814.3–39.4 | 214K | 9.7B | p46 | p24 | |
| Qwen/Qwen3.6-35B-A3B unsloth/Qwen3.6-35B-A3B-MTP-GGUF · UD-IQ3_XXS | Fits | 12951.7–142 | 256K | 36B | p43 | p41 | |
| openbmb/MiniCPM-V-4.7-35B-A3B openbmb/MiniCPM-V-4.7-35B-A3B-gguf · IQ3_XS | Fits | 12851.3–141 | 256K | 35.2B | p42 | – | |
| ornith-ai/Ornith-1.0-35B bartowski/deepreinforce-ai_Ornith-1.0-35B-GGUF · IQ3_XXS | Fits | 12750.7–139 | 256K | 35.1B | – | p40 | |
| openai/gpt-oss-20b unsloth/gpt-oss-20b-GGUF · F16 | Fits | 86.734.7–95.4 | 128K | 21.1B | p39 | p15 | |
| Qwen/Qwen3.5-4B unsloth/Qwen3.5-4B-GGUF · Q8_0 | Fits | 57.723.1–63.5 | 256K | 4.7B | p39 | – | |
| google/gemma-4-12B-it unsloth/gemma-4-12b-it-GGUF · Q8_0 | Fits | 25.310.1–27.9 | 154K | 12B | p38 | – | |
| Qwen/Qwen3-4B-Instruct-2507 BF16 | Fits | 31.912.7–35.1 | 52K | 4B | p38 | – | |
| Qwen/Qwen2.5-7B Mungert/Qwen2.5-7B-Instruct-1M-GGUF · Q8_0 | Fits | 38.015.2–41.8 | 128K | 7.6B | p29 | – | |
| google/gemma-4-E4B-it unsloth/gemma-4-E4B-it-GGUF · Q8_0 | Fits | 47.018.8–51.7 | 128K | 8B | p27 | – | |
| openbmb/MiniCPM5-2B BF16 | Fits | 55.022.0–60.5 | 128K | 2.5B | p27 | p3 | |
| poolside/Laguna-XS.2 poolside-laguna-hackathon/Piscina-XS.2-GGUF · IQ2 | Fits | 12550.1–138 | 256K | 33.4B | – | p25 | |
| Qwen/Qwen3.5-2B unsloth/Qwen3.5-2B-GGUF · BF16 | Fits | 68.727.5–75.6 | 256K | 2.3B | p24 | – | |
| ornith-ai/Ornith-1.0-9B unsloth/Ornith-1.0-9B-GGUF · Q8_0 | Fits | 35.914.4–39.5 | 214K | 9.4B | – | p20 | |
| google/gemma-4-E2B-it unsloth/gemma-4-E2B-it-GGUF · Q8_0 | Fits | 74.229.7–81.7 | 128K | 5.1B | p18 | – | |
| Qwen/Qwen2.5-1.5B BF16 | Fits | 73.729.5–81.1 | 128K | 1.5B | p15 | – | |
| Qwen/Qwen2.5-0.5B BF16 | Fits | 13855.1–152 | 32K | 0.5B | p5 | – | |
| Qwen/Qwen3.5-0.8B BF16 | Fits | 10943.7–120 | 256K | 0.9B | p4 | – | |
| Qwen/Qwen2.5-0.5B-Instruct BF16 | Fits | 13855.1–152 | 32K | 0.5B | – | – | |
| incoai/Qwen3.8-27B-DFlash2 BF16 | Fits | 13353.4–147 | 256K | 1.9B | – | – | |
| datalab-to/surya-ocr-2 BF16 | Fits | 12248.9–135 | 256K | 0.7B | – | – | |
| Qwen/Qwen3.5-0.8B-Base BF16 | Fits | 10943.7–120 | 256K | 0.9B | – | – | |
| baidu/Unlimited-OCR-2 AtomicChat/Unlimited-OCR-2-GGUF · BF16 | Fits | 10341.2–113 | 128K | 3.3B | – | – | |
| baidu/Unlimited-OCR sahilchachra/Unlimited-OCR-GGUF · BF16 | Fits | 95.638.2–105 | 32K | 3.3B | – | – | |
| Qwen/Qwen2.5-1.5B-Instruct Qwen/Qwen2.5-1.5B-Instruct-GGUF · Q8_0 | Fits | 95.438.2–105 | 32K | 1.5B | – | – | |
| openbmb/MiniCPM-V-4.7-1B BF16 | Fits | 88.135.2–96.9 | 256K | 1.6B | – | – | |
| Qwen/Qwen3-0.6B-Base BF16 | Fits | 82.933.1–91.2 | 32K | 0.6B | – | – | |
| Qwen/Qwen3-Coder-30B-A3B-Instruct unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF · Q3_K_S | Fits | 80.432.2–88.4 | 256K | 30.5B | – | – | |
| Qwen/Qwen3-30B-A3B unsloth/Qwen3-30B-A3B-GGUF · Q3_K_S | Fits | 80.432.2–88.4 | 40K | 30.5B | – | – | |
| openbmb/MiniCPM5-2B-GGUF Q8_0 | Fits | 78.531.4–86.3 | 128K | 2.5B | – | – | |
| Qwen/Qwen3-0.6B BF16 | Fits | 77.731.1–85.4 | 40K | 0.8B | – | – | |
| Qwen/Qwen2-VL-2B-Instruct ggml-org/Qwen2-VL-2B-Instruct-GGUF · F16 | Fits | 73.629.5–81.0 | 32K | 2.2B | – | – | |
| nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16 unsloth/NVIDIA-Nemotron-3-Nano-4B-GGUF · Q8_0 | Fits | 68.727.5–75.5 | 256K | 4B | – | – | |
| Qwen/Qwen2.5-3B-Instruct Qwen/Qwen2.5-3B-Instruct-GGUF · Q8_0 | Fits | 64.425.7–70.8 | 32K | 3.1B | – | – | |
| Qwen/Qwen3-VL-2B-Instruct Qwen/Qwen3-VL-2B-Instruct-GGUF · F16 | Fits | 55.922.3–61.4 | 106K | 2.1B | – | – | |
| Qwen/Qwen3-1.7B-Base BF16 | Fits | 55.922.4–61.5 | 32K | 1.7B | – | – | |
| Qwen/Qwen3-1.7B BF16 | Fits | 51.320.5–56.4 | 40K | 2B | – | – | |
| prism-ml/Bonsai-27B-mlx-1bit MLX-1BIT | Fits | 50.420.2–55.5 | 155K | 26.4B | – | – | |
| Qwen/Qwen2.5-VL-3B-Instruct unsloth/Qwen2.5-VL-3B-Instruct-GGUF · BF16 | Fits | 45.118.0–49.6 | 125K | 3.8B | – | – | |
| ornith-ai/Ornith-1.5-9B-NVFP4 MIXED_PRECISION | Fits | 38.815.5–42.7 | 204K | 9B | – | – | |
| Qwen/Qwen2.5-7B-Instruct Qwen/Qwen2.5-7B-Instruct-GGUF · Q8_0 | Fits | 38.015.2–41.8 | 32K | 7.6B | – | – | |
| Qwen/Qwen2.5-VL-7B-Instruct unsloth/Qwen2.5-VL-7B-Instruct-GGUF · Q8_0 | Fits | 37.815.1–41.6 | 125K | 8.3B | – | – | |
| Qwen/Qwen2-VL-7B-Instruct ggml-org/Qwen2-VL-7B-Instruct-GGUF · Q8_0 | Fits | 37.815.1–41.6 | 32K | 8.3B | – | – | |
| ornith-ai/Ornith-1.0-9B-GGUF Q8_0 | Fits | 36.114.4–39.7 | 214K | 9B | – | – | |
| prism-ml/Ternary-Bonsai-27B-mlx-2bit MLX-2BIT | Fits | 35.614.2–39.1 | 106K | 26.4B | – | – | |
| ornith-ai/Ornith-1.5-9B-GGUF Q8_0 | Fits | 35.214.1–38.8 | 207K | 9.2B | – | – | |
| Qwen/Qwen2.5-Coder-7B-Instruct Qwen/Qwen2.5-Coder-7B-Instruct-GGUF · Q4_K_M | Fits | 34.113.6–37.5 | 32K | 7.6B | – | – | |
| HauhauCS/Gemma-4-E4B-Uncensored-HauhauCS-Aggressive Q8_K_P | Fits | 34.013.6–37.4 | 128K | 7.5B | – | – | |
| Qwen/Qwen3.5-4B-Base BF16 | Fits | 33.713.5–37.1 | 189K | 4.7B | – | – | |
| Qwen/Qwen3-8B-Base Qwen/Qwen3-Embedding-8B-GGUF · Q8_0 | Fits | 33.613.4–36.9 | 32K | 8.2B | – | – | |
| Qwen/Qwen3-8B unsloth/Qwen3-8B-GGUF · Q8_0 | Fits | 31.912.7–35.0 | 40K | 8.2B | – | – | |
| Qwen/Qwen3-VL-4B-Instruct Qwen/Qwen3-VL-4B-Instruct-GGUF · F16 | Fits | 31.912.7–35.0 | 52K | 4.4B | – | – | |
| Qwen/Qwen3-4B BF16 | Fits | 31.912.7–35.1 | 40K | 4B | – | – | |
| Qwen/Qwen3-VL-8B-Instruct Qwen/Qwen3-VL-8B-Instruct-GGUF · Q8_0 | Fits | 31.712.7–34.9 | 52K | 8.8B | – | – | |
| datalab-to/chandra-ocr-2 BF16 | Fits | 30.412.1–33.4 | 153K | 5.3B | – | – | |
| Qwen/Qwen3-32B unsloth/Qwen3-32B-GGUF · UD-IQ2_XXS | Fits | 25.010.0–27.5 | 40K | 32.8B | – | – | |
| Qwen/Qwen3-14B unsloth/Qwen3-14B-GGUF · Q6_K | Fits | 24.29.70–26.6 | 40K | 14.8B | – | – | |
| Qwen/Qwen2.5-14B-Instruct Qwen/Qwen2.5-14B-Instruct-GGUF · Q6_K | Fits | 23.49.30–25.7 | 32K | 14.8B | – | – | |
| Qwen/Qwen2.5-Coder-14B-Instruct bartowski/Qwen2.5-Coder-14B-Instruct-GGUF · Q6_K_L | Fits | 22.99.10–25.1 | 32K | 14.8B | – | – | |
| Qwen/Qwen2.5-32B-Instruct bartowski/Qwen2.5-32B-Instruct-GGUF · IQ2_M | Fits | 22.49.00–24.7 | 32K | 32.8B | – | – | |
| Qwen/Qwen2.5-Coder-32B-Instruct bartowski/Qwen2.5-Coder-32B-Instruct-GGUF · IQ2_M | Fits | 22.49.00–24.7 | 32K | 32.8B | – | – | |
| farbodtavakkoli/OTel-2.0-LLM-31B-IT wfakhri/OTel-2.0-LLM-31B-IT-GGUF · IQ3_XS | Fits | 20.58.20–22.5 | 41K | 31.3B | – | – | |
| zai-org/GLM-5.3 unsloth/GLM-5.3-GGUF · Q8_0 | Offload | 3.701.90–5.60 | 0 | 753.3B | p81 | p58 | |
| Qwen/Qwen3.5-122B-A10B unsloth/Qwen3.5-122B-A10B-GGUF · Q4_K_S | Offload | 16.46.60–24.6 | 165K | 125.1B | p77 | p49 | |
| zai-org/GLM-5 unsloth/GLM-5-GGUF · Q5_K_M | Offload | 12.36.10–18.4 | 0 | 753.9B | p74 | p71 | |
| tencent/Hy3 bartowski/Hy3-GGUF · Q4_K_M | Offload | 19.49.70–29.1 | 0 | 298.8B | p73 | p46 | |
| zai-org/GLM-5.1 bartowski/zai-org_GLM-5.1-GGUF · Q5_K_M | Offload | 11.65.80–17.5 | 0 | 753.9B | p53 | p68 | |
| Qwen/Qwen3.8-Flash-Next ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF · IQ2_XS | Offload | 25.910.4–38.9 | 256K | 180B | p68 | p51 | |
| thinkingmachines/Inkling DevQuasar/thinkingmachines.Inkling-GGUF · Q4_K_M | Offload | 5.002.50–7.40 | 0 | 952.4B | p68 | p45 | |
| nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 unsloth/NVIDIA-Nemotron-3-Ultra-550B-A55B-GGUF · BF16 | Offload | 6.002.40–9.10 | 0 | 560.5B | p62 | p33 | |
| deepseek-ai/DeepSeek-V3.2 unsloth/DeepSeek-V3.2-GGUF · Q6_K | Offload | 7.103.60–10.7 | 0 | 685.4B | p60 | p34 | |
| thinkingmachines/Inkling-Small unsloth/Inkling-Small-GGUF · BF16 | Offload | 7.803.90–11.6 | 0 | 266B | p57 | p52 | |
| nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16 bartowski/nvidia_Nemotron-3-Super-120B-A12B-GGUF · IQ4_NL | Offload | 13.95.60–20.9 | 256K | 123.6B | p50 | p15 | |
| openai/gpt-oss-120b unsloth/gpt-oss-120b-GGUF · F16 | Offload | 18.89.40–28.2 | 128K | 117B | p47 | p15 | |
| poolside/Laguna-M.1 sigargv/Laguna-M.1-GGUF · Q8_0 | Offload | 10.55.20–15.7 | 0 | 225.8B | – | p45 | |
| nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF · Q8_0 | Offload | 28.611.4–42.8 | 256K | 32.9B | p37 | p9 | |
| nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 MIXED_PRECISION | Offload | 57.122.8–85.7 | 1M | 47.1B | p36 | p9 | |
| poolside/Laguna-S-2.1 Myric/Laguna-S-2.1-APEX-GGUF · V2 [quality] | Offload | 13.25.30–19.8 | 63K | 117.6B | – | p34 | |
| nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 unsloth/Nemotron-3-Nano-30B-A3B-GGUF · Q5_K_S | Offload | 48.619.4–72.9 | 256K | 31.6B | p33 | – | |
| poolside/Laguna-XS-2.1 poolside/Laguna-XS-2.1-GGUF · BF16 | Offload | 17.48.70–26.0 | 215K | 33.4B | – | p30 | |
| nvidia/Qwen3.6-35B-A3B-NVFP4 MIXED_PRECISION | Offload | 61.924.8–92.8 | 256K | 34B | – | – | |
| ornith-ai/Ornith-1.5-35B-A3B-NVFP4 MIXED_PRECISION | Offload | 61.924.8–92.9 | 256K | 34B | – | – | |
| ornith-ai/Ornith-1.0-35B-GGUF Q8_0 | Offload | 32.613.0–48.9 | 256K | 34.7B | – | – | |
| ornith-ai/Ornith-1.5-35B-A3B-GGUF Q8_0 | Offload | 31.712.7–47.6 | 256K | 35.5B | – | – | |
| ornith-ai/Ornith-1.5-35B-A3B-FP8 COMPRESSED-TENSORS | Offload | 27.511.0–41.3 | 256K | 36B | – | – | |
| Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 FP8 | Offload | 21.68.60–32.4 | 136K | 30.5B | – | – |
Quality = benchmark percentile against all indexed models (p50 = median). Speeds are single-user decode ranges.