Your rig
M3 Ultra · 512 GB · 512 GB
512 GB unified · 800 GB/s · 384.0 GB addressable by the GPU
Which one should I actually run?
Best models for your M3 Ultra · 512 GB
What do you want to do? Each row is the best-ranked model on that use case's shortlist that fits at Q2_K and 128K. The ordering is editorial; the sizes and speeds are computed. How the shortlists work.
All
General assistantCodingReasoning & mathsWritingVisionAgents & tool useTranslationLow latencyLong context
| Recommendation | Model | Why | Quant · total | Tok/s est. | |
|---|---|---|---|---|---|
| Best overall | GLM-5.2 744B-A40B Z.ai · MIT |
The strongest all-round open model on the mid-2026 boards — if you have the memory for 744B of weights. | Q2_K · 305.2 GB | 13 | Runs great |
| Best for coding | GLM-5.2 744B-A40B Z.ai · MIT |
Top-tier on SWE-bench Pro and terminal agents, MIT-licensed. | Q2_K · 305.2 GB | 13 | Runs great |
| Best reasoning | GLM-5.2 744B-A40B Z.ai · MIT |
Frontier-class reasoning, MIT-licensed, 744B. | Q2_K · 305.2 GB | 13 | Runs great |
| Best for writing | Llama 3.3 70B Instruct Meta · Llama 3.3 Community |
The creative-writing favourite: consistent voice, takes direction, handles fiction framing. | Q2_K · 68.1 GB | 16 | Runs great |
| Best vision | Qwen3.8 27B New this week Alibaba · Apache 2.0 |
Image and video in, OSWorld 84 — the strongest local vision-language model at 24 GB. | Q2_K · 19.4 GB | 41 | Runs great |
| Best for agents | GLM-5.2 744B-A40B Z.ai · MIT |
Frontier-class tool loops, MIT-licensed. | Q2_K · 305.2 GB | 13 | Runs great |
| Best translation | Qwen3.8 27B New this week Alibaba · Apache 2.0 |
119 languages, and Qwen translations read as idiomatic rather than literal. | Q2_K · 19.4 GB | 41 | Runs great |
| Fastest good model | Qwen3.6 35B-A3B Alibaba · Apache 2.0 |
3.3B active per token: the fastest model that still competes with dense 27Bs. | Q2_K · 17.1 GB | 162 | Runs great |
| Best long context | DeepSeek V4 Flash 284B-A13B New DeepSeek · MIT |
A 1M-token window with a compressed latent cache. | Q2_K · 117.4 GB at 128K context |
41 | Runs great |
Everything that fits
Runs great
72 of 76 · fits with headroom for a longer prompt| Model | Params | Quant | Weights | +KV 128K | Total | VRAM headroom | Tok/s | Runtime | |
|---|---|---|---|---|---|---|---|---|---|
| GLM-5.2 744B-A40B | 753B MoE / 40B act | Q2_K | 293.7 GB | 10.97 GB | 305.2 GB |
78.8 GB free
|
13 | MLX | Run |
| DeepSeek-R1 671B | 671B MoE / 37B act | Q2_K | 261.7 GB | 8.58 GB | 270.9 GB |
113.1 GB free
|
14 | MLX | Run |
| Qwen3.5 397B-A17B | 403B MoE / 17B act | Q2_K | 157.2 GB | 3.75 GB | 161.5 GB |
222.5 GB free
|
31 | MLX | Run |
| Ornith 1.5 397B-A17B New this week | 397B MoE / 17B act | Q2_K | 154.8 GB | 3.75 GB | 159.2 GB |
224.8 GB free
|
31 | MLX | Run |
| DeepSeek V4 Flash 284B-A13B New | 284B MoE / 13B act | Q2_K | 110.8 GB | 6.05 GB | 117.4 GB |
266.6 GB free
|
41 | MLX | Run |
| Qwen3 235B-A22B | 235B MoE / 22B act | Q2_K | 91.6 GB | 23.50 GB | 115.7 GB |
268.3 GB free
|
24 | MLX | Run |
| Qwen3.5 122B-A10B | 125B MoE / 10B act | Q2_K | 48.7 GB | 3.00 GB | 52.3 GB |
331.7 GB free
|
53 | MLX | Run |
| Nemotron 3 Super 120B-A12B | 124B MoE / 12B act | Q2_K | 48.4 GB | 1.00 GB | 50.0 GB |
334.0 GB free
|
45 | MLX | Run |
| Ling 3.0 Flash 124B-A5B New | 124B MoE / 5.1B act | Q2_K | 48.4 GB | 0.98 GB | 49.9 GB |
334.1 GB free
|
105 | MLX | Run |
| Mistral Small 4 119B-A6B | 119B MoE / 6.5B act | Q2_K | 46.4 GB | 2.81 GB | 49.8 GB |
334.2 GB free
|
82 | MLX | Run |
| gpt-oss 120B | 117B MoE / 5.1B act | MXFP4 | 60.5 GB | 4.50 GB | 65.6 GB |
318.4 GB free
|
79 | MLX | Run |
| Llama 4 Scout 109B-A17B | 109B MoE / 17B act | Q2_K | 42.5 GB | 7.13 GB | 50.2 GB |
333.8 GB free
|
31 | MLX | Run |
| Llama 3.3 70B Instruct | 70.6B | Q2_K | 27.5 GB | 40.00 GB | 68.1 GB |
315.9 GB free
|
16 | MLX | Run |
| DeepSeek-R1-Distill-Llama 70B | 70.6B | Q2_K | 27.5 GB | 40.00 GB | 68.1 GB |
315.9 GB free
|
16 | MLX | Run |
| Qwen3.6 35B-A3B | 35.9B MoE / 3.3B act | Q2_K | 14.0 GB | 2.50 GB | 17.1 GB |
366.9 GB free
|
162 | MLX | Run |
| Ornith 1.5 35B-A3B New this week | 35.9B MoE / 3.3B act | Q2_K | 14.0 GB | 2.50 GB | 17.1 GB |
366.9 GB free
|
162 | MLX | Run |
| LLM-jp 4 33B Thinking New this week | 33.2B | Q2_K | 12.9 GB | 16.00 GB | 29.5 GB |
354.5 GB free
|
35 | MLX | Run |
| Qwen3 32B | 32.8B | Q2_K | 12.8 GB | 32.00 GB | 45.4 GB |
338.6 GB free
|
35 | MLX | Run |
| DeepSeek-R1-Distill-Qwen 32B | 32.8B | Q2_K | 12.8 GB | 32.00 GB | 45.4 GB |
338.6 GB free
|
35 | MLX | Run |
| Qwen2.5-Coder 32B | 32.8B | Q2_K | 12.8 GB | 32.00 GB | 45.4 GB |
338.6 GB free
|
35 | MLX | Run |
| Olmo 3.1 32B Instruct | 32.2B | Q2_K | 12.6 GB | 4.75 GB | 17.9 GB |
366.1 GB free
|
36 | MLX | Run |
| Nemotron 3.5 Lightning 30B-A3B New | 31.6B MoE / 3.2B act | Q2_K | 12.3 GB | 0.75 GB | 13.7 GB |
370.3 GB free
|
167 | MLX | Run |
| Gemma 4 31B | 31.3B | Q2_K | 12.2 GB | 20.78 GB | 33.6 GB |
350.4 GB free
|
37 | MLX | Run |
| GLM-4.7-Flash 30B-A3B | 31.2B MoE / 3B act | Q2_K | 12.2 GB | 6.61 GB | 19.4 GB |
364.6 GB free
|
178 | MLX | Run |
| Qwen3 30B-A3B | 30.5B MoE / 3.3B act | Q2_K | 11.9 GB | 12.00 GB | 24.5 GB |
359.5 GB free
|
162 | MLX | Run |
| Qwen3 Coder 30B-A3B | 30.5B MoE / 3.3B act | Q2_K | 11.9 GB | 12.00 GB | 24.5 GB |
359.5 GB free
|
162 | MLX | Run |
| Muse Glimmer 30B New | 29.8B | Q2_K | 11.6 GB | 1.70 GB | 13.9 GB |
370.1 GB free
|
38 | MLX | Run |
| Granite 4.1 30B | 28.9B | Q2_K | 11.3 GB | 32.00 GB | 43.9 GB |
340.1 GB free
|
40 | MLX | Run |
| Qwen3.8 27B New this week | 27.8B | Q2_K | 10.8 GB | 8.00 GB | 19.4 GB |
364.6 GB free
|
41 | MLX | Run |
| Qwen3.6 27B | 27.8B | Q2_K | 10.8 GB | 8.00 GB | 19.4 GB |
364.6 GB free
|
41 | MLX | Run |
| Gemma 3 27B | 27.4B | Q2_K | 10.7 GB | 10.41 GB | 21.7 GB |
362.3 GB free
|
42 | MLX | Run |
| Gemma 4 26B-A4B | 26.5B MoE / 3.8B act | Q2_K | 10.3 GB | 5.20 GB | 16.1 GB |
367.9 GB free
|
141 | MLX | Run |
| Devstral Small 2 24B | 24B | Q2_K | 9.4 GB | 20.00 GB | 30.0 GB |
354.0 GB free
|
48 | MLX | Run |
| Mistral Small 3.2 24B | 23.6B | Q2_K | 9.2 GB | 20.00 GB | 29.8 GB |
354.2 GB free
|
49 | MLX | Run |
| gpt-oss 20B | 20.9B MoE / 3.6B act | MXFP4 | 10.8 GB | 3.00 GB | 14.4 GB |
369.6 GB free
|
112 | MLX | Run |
| Qwen3 14B | 14.8B | Q2_K | 5.8 GB | 20.00 GB | 26.4 GB |
357.6 GB free
|
77 | MLX | Run |
| DeepSeek-R1-Distill-Qwen 14B | 14.8B | Q2_K | 5.8 GB | 24.00 GB | 30.4 GB |
353.6 GB free
|
77 | MLX | Run |
| Qwen2.5-Coder 14B | 14.8B | Q2_K | 5.8 GB | 24.00 GB | 30.4 GB |
353.6 GB free
|
77 | MLX | Run |
| Phi-4 14B | 14.7B | Q2_K | 5.7 GB | 3.13 GB | 9.5 GB |
374.5 GB free
|
78 | MLX | Run |
| Ministral 3 14B | 13.9B | Q2_K | 5.4 GB | 20.00 GB | 26.0 GB |
358.0 GB free
|
82 | MLX | Run |
| Gemma 3 12B | 12.2B | Q2_K | 4.8 GB | 8.31 GB | 13.7 GB |
370.3 GB free
|
94 | MLX | Run |
| Mistral NeMo 12B | 12.2B | Q2_K | 4.8 GB | 20.00 GB | 25.4 GB |
358.6 GB free
|
94 | MLX | Run |
| Gemma 4 12B | 12B | Q2_K | 4.7 GB | 8.31 GB | 13.6 GB |
370.4 GB free
|
96 | MLX | Run |
| Qwen3.5 9B | 9.65B | Q2_K | 3.8 GB | 4.00 GB | 8.4 GB |
375.6 GB free
|
119 | MLX | Run |
| Ornith 1.5 9B New this week | 9.41B | Q2_K | 3.7 GB | 4.00 GB | 8.3 GB |
375.7 GB free
|
122 | MLX | Run |
| Ministral 3 8B | 8.92B | Q2_K | 3.5 GB | 17.00 GB | 21.1 GB |
362.9 GB free
|
129 | MLX | Run |
| Granite 4.1 8B | 8.79B | Q2_K | 3.4 GB | 20.00 GB | 24.0 GB |
360.0 GB free
|
130 | MLX | Run |
| LFM2.5 8B-A1B | 8.47B MoE / 1.5B act | Q2_K | 3.3 GB | 1.46 GB | 5.4 GB |
378.6 GB free
|
357 | MLX | Run |
| Fara 7B | 8.29B | Q2_K | 3.2 GB | 6.84 GB | 10.7 GB |
373.3 GB free
|
138 | MLX | Run |
| Qwen3 8B | 8.19B | Q2_K | 3.2 GB | 18.00 GB | 21.8 GB |
362.2 GB free
|
140 | MLX | Run |
| Llama 3.1 8B Instruct | 8.03B | Q2_K | 3.1 GB | 16.00 GB | 19.7 GB |
364.3 GB free
|
143 | MLX | Run |
| Gemma 4 E4B | 8.0B | Q2_K | 3.1 GB | 1.78 GB | 5.5 GB |
378.5 GB free
|
143 | MLX | Run |
| Ling 3.0 Tiny 7.9B-A1.3B New | 7.9B MoE / 1.3B act | Q2_K | 3.1 GB | 0.84 GB | 4.5 GB |
379.5 GB free
|
411 | MLX | Run |
| DeepSeek-R1-Distill-Qwen 7B | 7.62B | Q2_K | 3.0 GB | 7.00 GB | 10.6 GB |
373.4 GB free
|
150 | MLX | Run |
| Qwen2.5-Coder 7B | 7.62B | Q2_K | 3.0 GB | 7.00 GB | 10.6 GB |
373.4 GB free
|
150 | MLX | Run |
| Olmo 3 7B Instruct | 7.3B | Q2_K | 2.8 GB | 9.50 GB | 12.9 GB |
371.1 GB free
|
157 | MLX | Run |
| Mistral 7B Instruct v0.3 | 7.25B | Q2_K | 2.8 GB | 4.00 GB | 7.4 GB |
376.6 GB free
|
158 | MLX | Run |
| Gemma 4 E2B | 5.1B | Q2_K | 2.0 GB | 0.89 GB | 3.5 GB |
380.5 GB free
|
225 | MLX | Run |
| Qwen3.5 4B | 4.66B | Q2_K | 1.8 GB | 4.00 GB | 6.4 GB |
377.6 GB free
|
246 | MLX | Run |
| Gemma 3 4B | 4.3B | Q2_K | 1.7 GB | 3.11 GB | 5.4 GB |
378.6 GB free
|
267 | MLX | Run |
| Qwen3 4B | 4.02B | Q2_K | 1.6 GB | 4.50 GB | 6.7 GB |
377.3 GB free
|
285 | MLX | Run |
| Ministral 3 3B | 3.85B | Q2_K | 1.5 GB | 13.00 GB | 15.1 GB |
368.9 GB free
|
298 | MLX | Run |
| Phi-4-mini 3.8B | 3.84B | Q2_K | 1.5 GB | 16.00 GB | 18.1 GB |
365.9 GB free
|
299 | MLX | Run |
| Granite 4.1 3B | 3.4B | Q2_K | 1.3 GB | 10.00 GB | 11.9 GB |
372.1 GB free
|
337 | MLX | Run |
| Llama 3.2 3B Instruct | 3.21B | Q2_K | 1.3 GB | 14.00 GB | 15.9 GB |
368.1 GB free
|
357 | MLX | Run |
| LFM2.5 2.6B New | 2.7B | Q2_K | 1.1 GB | 2.00 GB | 3.7 GB |
380.3 GB free
|
425 | MLX | Run |
| Qwen3.5 2B | 2.27B | Q2_K | 0.9 GB | 1.50 GB | 3.0 GB |
381.0 GB free
|
505 | MLX | Run |
| Qwen3 1.7B | 1.72B | Q2_K | 0.7 GB | 3.50 GB | 4.8 GB |
379.2 GB free
|
666 | MLX | Run |
| Llama 3.2 1B Instruct | 1.24B | Q2_K | 0.5 GB | 4.00 GB | 5.1 GB |
378.9 GB free
|
924 | MLX | Run |
| Gemma 3 1B | 1.0B | Q2_K | 0.4 GB | 0.14 GB | 1.1 GB |
382.9 GB free
|
1146 | MLX | Run |
| Qwen3.5 0.8B | 0.87B | Q2_K | 0.3 GB | 1.50 GB | 2.4 GB |
381.6 GB free
|
1318 | MLX | Run |
| Qwen3 0.6B | 0.6B | Q2_K | 0.2 GB | 3.50 GB | 4.3 GB |
379.7 GB free
|
1910 | MLX | Run |
Won't fit in VRAM
4 models · some are possible with CPU offload| Model | Params | Quant | Weights | +KV 128K | Total | Over budget | Tok/s | Runtime | |
|---|---|---|---|---|---|---|---|---|---|
| Kimi K3 2.8T-A104B New | 2780B MoE / 104B act | Q2_K | 1084.2 GB | 3.38 GB | 1088.2 GB | +704.2 GB | ~5.1 | MLX | Why |
| Qwen3.8 2.4T-A95B New | 2446B MoE / 95B act | Q2_K | 953.9 GB | 11.50 GB | 966.0 GB | +582.0 GB | ~5.6 | MLX | Why |
| DeepSeek V4 Pro 1.6T-A49B New | 1650B MoE / 49B act | Q2_K | 643.5 GB | 8.58 GB | 652.7 GB | +268.7 GB | ~11 | MLX | Why |
| Kimi K2.6 1T-A32B | 1027B MoE / 32B act | Q2_K | 400.5 GB | 8.58 GB | 409.7 GB | +25.7 GB | ~17 | MLX | Why |
Totals = quantised weights + f16 KV cache at 128K tokens + 0.6 GB runtime overhead, against 384.0 GB usable unified memory. Tokens per second are estimated from 800 GB/s peak memory bandwidth at batch 1 — not measured. Read the method.