Model directory
Which LLMs can your computer run?
All 69 tracked open models, ordered by current Arena Text Overall, style-controlled reference rating, with unrated models afterward — with the memory each needs at its smallest quantization and how many of our tracked machines fit it. Open a model to check your exact computer, or start from your machine instead.
Snapshot published 2026-10-08 · snapshot imported 2026-10-11 UTC
Validated last-good snapshot · manual refresh. Reference ratings remain available.
Practical model setup guides
Compare memory, context and runtime choices before downloading a model.
Tracked models
Arena reference · minimum total memory
- Hunyuan Hy3 (MoE)Hunyuan1,454≥ 108.8 GB10/88 machines
- Gemma 4 31BGemma1,452≥ 13.7 GB59/88 machines
- Qwen3.8 27BQwen1,438≥ 11.9 GB61/88 machines
- Gemma 4 26B-A4B (MoE)Gemma1,438≥ 11.6 GB61/88 machines
- Qwen3 30B-A3B (MoE)Qwen1,382≥ 12.5 GB61/88 machines
- GLM-4.5 Air (MoE)GLM1,373≥ 46.9 GB21/88 machines
- Gemma 3 27BGemma1,365≥ 12.1 GB61/88 machines
- GLM-4.7 Flash (MoE)GLM1,365≥ 12.4 GB61/88 machines
- Mistral Small 3.2 24BMistral1,356≥ 10.4 GB72/88 machines
- GPT-OSS 120B (MoE)GPT-OSS1,352≥ 63.5 GB21/88 machines
- Qwen3 32BQwen1,347≥ 14.2 GB59/88 machines
- Gemma 3 12BGemma1,342≥ 6.2 GB81/88 machines
- Granite 4.2 30BGranite1,340≥ 12.7 GB60/88 machines
- Llama 4 Scout (MoE)Llama1,322≥ 41.2 GB30/88 machines
- Llama 3.3 70B InstructLlama1,318≥ 28.5 GB34/88 machines
- GPT-OSS 20B (MoE)GPT-OSS1,317≥ 12.4 GB61/88 machines
- Granite 4.2 8BGranite1,290≥ 4.9 GB87/88 machines
- Phi-4 14BPhi1,256≥ 7.2 GB73/88 machines
- Llama 3.1 8BLlama1,211≥ 4.5 GB87/88 machines
- Gemma 2 2BGemma1,200≥ 2.9 GB88/88 machines
- Llama 3.2 3BLlama1,167≥ 3.3 GB88/88 machines
- Llama 3.2 1BLlama1,111≥ 1.7 GB88/88 machines
- 1-bit Bonsai 27BBonsaiUnrated≥ 4.9 GB87/88 machines
- AesCode 32B (text-only)AesCodeUnrated≥ 14.9 GB52/88 machines
- AesCode 8B (text-only)AesCodeUnrated≥ 4.7 GB87/88 machines
- DeepSeek V4 Flash (MoE)DeepSeekUnrated≥ 98.0 GB10/88 machines
- DeepSeek-R1-Distill 14BDeepSeekUnrated≥ 7.4 GB73/88 machines
- DeepSeek-R1-Distill 32BDeepSeekUnrated≥ 14.2 GB59/88 machines
- DeepSeek-R1-Distill 8BDeepSeekUnrated≥ 4.5 GB87/88 machines
- Devstral Small 2 24BMistralUnrated≥ 10.4 GB72/88 machines
- Gemma 4 12BGemmaUnrated≥ 5.9 GB81/88 machines
- Gemma 4 E4BGemmaUnrated≥ 4.6 GB87/88 machines
- GLM-4 9BGLMUnrated≥ 5.0 GB87/88 machines
- GLM-5.2 (MoE)GLMUnrated≥ 255.1 GB5/88 machines
- GLM-5.3 (MoE)GLMUnrated≥ 255.1 GB5/88 machines
- GLM-5.3 Flash (MoE)GLMUnrated≥ 126.6 GB10/88 machines
- Kimi K2.7 (MoE)KimiUnrated≥ 340.5 GB4/88 machines
- Kimi K3 (MoE)KimiUnrated≥ 862.2 GB0/88 machines
- Kimi Linear 48B-A3B (MoE)KimiUnrated≥ 18.4 GB46/88 machines
- Laguna S 2.1 (MoE)LagunaUnrated≥ 40.8 GB30/88 machines
- Laguna XS 2.1 (MoE)LagunaUnrated≥ 13.2 GB59/88 machines
- LFM2.5 1.2BLFMUnrated≥ 1.6 GB88/88 machines
- LFM2.5 2.6BLFMUnrated≥ 2.5 GB88/88 machines
- LFM2.5 24B A2BLFMUnrated≥ 8.9 GB73/88 machines
- LFM2.5 8B-A1B (MoE)LFMUnrated≥ 6.0 GB81/88 machines
- Ling 3.0 Flash (MoE)LingUnrated≥ 46.7 GB21/88 machines
- Mellum2.1 12B A2.5B Thinking (MoE)MellumUnrated≥ 9.0 GB73/88 machines
- MiMo V2.6 Distill Qwen 9BMiMoUnrated≥ 4.6 GB87/88 machines
- MiniCPM5 2BMiniCPMUnrated≥ 2.0 GB88/88 machines
- Mistral Small 4 119B (MoE)MistralUnrated≥ 41.1 GB30/88 machines
- Muse Glimmer 30BMuseUnrated≥ 12.0 GB61/88 machines
- Nemotron 3.5 Lightning 30B-A3B (MoE)NemotronUnrated≥ 19.7 GB46/88 machines
- North Mini Code 1.0 (MoE)NorthUnrated≥ 11.7 GB61/88 machines
- NVIDIA Nemotron 3.5 Bolt 74B A9BNemotronUnrated≥ 29.6 GB33/88 machines
- Ornith 1.0 35B (MoE)OrnithUnrated≥ 13.1 GB59/88 machines
- Ornith 1.0 9BOrnithUnrated≥ 6.6 GB81/88 machines
- Ornith 1.5 35B-A3B (MoE)OrnithUnrated≥ 14.0 GB59/88 machines
- Ornith 1.5 9BOrnithUnrated≥ 5.0 GB87/88 machines
- Qwen3 14BQwenUnrated≥ 7.2 GB73/88 machines
- Qwen3 8BQwenUnrated≥ 4.7 GB87/88 machines
- Qwen3 Coder 30B-A3B (MoE)QwenUnrated≥ 12.5 GB61/88 machines
- Qwen3 Coder Next 80B-A3B (MoE)QwenUnrated≥ 30.1 GB33/88 machines
- Qwen3.5 4BQwenUnrated≥ 2.9 GB88/88 machines
- Qwen3.5 9BQwenUnrated≥ 5.1 GB87/88 machines
- Qwen3.6 27BQwenUnrated≥ 12.9 GB59/88 machines
- Qwen3.6 35B-A3B (MoE)QwenUnrated≥ 13.2 GB59/88 machines
- Qwen3.8 Flash Next (MoE)QwenUnrated≥ 81.8 GB16/88 machines
- Step 3.7 Flash (MoE)StepUnrated≥ 72.3 GB16/88 machines
- Ternary Bonsai 2 27BBonsaiUnrated≥ 7.0 GB81/88 machines
Arena reference ratings keep their published units and do not measure local quantization. Unrated models remain usable. Memory = smallest-quant footprint at 4K context · machine counts across the tracked device catalog. Source and mapping methodology.