Your rig
M3 Ultra · 256 GB · 256 GB
256 GB unified · 800 GB/s · 192.0 GB addressable by the GPU
Which one should I actually run?
Best models for your M3 Ultra · 256 GB
What do you want to do? Each row is the best-ranked model on that use case's shortlist that fits at Q6_K and 8K. 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 | Qwen3.8 27B New this week Alibaba · Apache 2.0 |
Best capability per gigabyte: a current-generation dense 27B with vision and 262K context. | Q6_K · 22.3 GB | 21 | Runs great |
| Best for coding | Ling 3.0 Flash 124B-A5B New inclusionAI · MIT |
SWE-bench Pro 56.6 claimed at 5B active — the 128 GB-class coding pick if you cannot fit DeepSeek V4. | Q6_K · 95.4 GB | 54 | Runs great |
| Best reasoning | Ling 3.0 Flash 124B-A5B New inclusionAI · MIT |
AIME 93 claimed at 5B active; hybrid attention keeps long reasoning traces cheap. | Q6_K · 95.4 GB | 54 | 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. | Q6_K · 57.0 GB | 8.3 | 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. | Q6_K · 22.3 GB | 21 | Runs great |
| Best for agents | Ling 3.0 Flash 124B-A5B New inclusionAI · MIT |
Agentic workflows are what Ling 3.0 was trained for; MIT, 5B active. | Q6_K · 95.4 GB | 54 | 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. | Q6_K · 22.3 GB | 21 | 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. | Q6_K · 28.2 GB | 83 | Runs great |
| Best long context | Qwen3.8 27B New this week Alibaba · Apache 2.0 |
262K native, and only 16 of 64 blocks keep a KV cache — long context is cheap here. | Q6_K · 29.8 GB at 128K context |
21 | Runs great |
Everything that fits
Runs great
66 of 76 · fits with headroom for a longer prompt| Model | Params | Quant | Weights | +KV 8K | Total | VRAM headroom | Tok/s | Runtime | |
|---|---|---|---|---|---|---|---|---|---|
| Qwen3.5 122B-A10B | 125B MoE / 10B act | Q6_K | 95.5 GB | 0.19 GB | 96.2 GB |
95.8 GB free
|
27 | MLX | Run |
| Nemotron 3 Super 120B-A12B | 124B MoE / 12B act | Q6_K | 94.7 GB | 0.06 GB | 95.4 GB |
96.6 GB free
|
23 | MLX | Run |
| Ling 3.0 Flash 124B-A5B New | 124B MoE / 5.1B act | Q6_K | 94.7 GB | 0.06 GB | 95.4 GB |
96.6 GB free
|
54 | MLX | Run |
| Mistral Small 4 119B-A6B | 119B MoE / 6.5B act | Q6_K | 90.9 GB | 0.18 GB | 91.7 GB |
100.3 GB free
|
42 | MLX | Run |
| gpt-oss 120B | 117B MoE / 5.1B act | MXFP4 | 60.5 GB | 0.29 GB | 61.4 GB |
130.6 GB free
|
79 | MLX | Run |
| Llama 4 Scout 109B-A17B | 109B MoE / 17B act | Q6_K | 83.2 GB | 1.50 GB | 85.3 GB |
106.7 GB free
|
16 | MLX | Run |
| Llama 3.3 70B Instruct | 70.6B | Q6_K | 53.9 GB | 2.50 GB | 57.0 GB |
135.0 GB free
|
8.3 | MLX | Run |
| DeepSeek-R1-Distill-Llama 70B | 70.6B | Q6_K | 53.9 GB | 2.50 GB | 57.0 GB |
135.0 GB free
|
8.3 | MLX | Run |
| Qwen3.6 35B-A3B | 35.9B MoE / 3.3B act | Q6_K | 27.4 GB | 0.16 GB | 28.2 GB |
163.8 GB free
|
83 | MLX | Run |
| Ornith 1.5 35B-A3B New this week | 35.9B MoE / 3.3B act | Q6_K | 27.4 GB | 0.16 GB | 28.2 GB |
163.8 GB free
|
83 | MLX | Run |
| LLM-jp 4 33B Thinking New this week | 33.2B | Q6_K | 25.4 GB | 2.00 GB | 28.0 GB |
164.0 GB free
|
18 | MLX | Run |
| Qwen3 32B | 32.8B | Q6_K | 25.0 GB | 2.00 GB | 27.6 GB |
164.4 GB free
|
18 | MLX | Run |
| DeepSeek-R1-Distill-Qwen 32B | 32.8B | Q6_K | 25.0 GB | 2.00 GB | 27.6 GB |
164.4 GB free
|
18 | MLX | Run |
| Qwen2.5-Coder 32B | 32.8B | Q6_K | 25.0 GB | 2.00 GB | 27.6 GB |
164.4 GB free
|
18 | MLX | Run |
| Olmo 3.1 32B Instruct | 32.2B | Q6_K | 24.6 GB | 1.25 GB | 26.4 GB |
165.6 GB free
|
18 | MLX | Run |
| Nemotron 3.5 Lightning 30B-A3B New | 31.6B MoE / 3.2B act | Q6_K | 24.1 GB | 0.05 GB | 24.8 GB |
167.2 GB free
|
85 | MLX | Run |
| Gemma 4 31B | 31.3B | Q6_K | 23.9 GB | 2.03 GB | 26.5 GB |
165.5 GB free
|
19 | MLX | Run |
| GLM-4.7-Flash 30B-A3B | 31.2B MoE / 3B act | Q6_K | 23.8 GB | 0.41 GB | 24.8 GB |
167.2 GB free
|
91 | MLX | Run |
| Qwen3 30B-A3B | 30.5B MoE / 3.3B act | Q6_K | 23.3 GB | 0.75 GB | 24.6 GB |
167.4 GB free
|
83 | MLX | Run |
| Qwen3 Coder 30B-A3B | 30.5B MoE / 3.3B act | Q6_K | 23.3 GB | 0.75 GB | 24.6 GB |
167.4 GB free
|
83 | MLX | Run |
| Muse Glimmer 30B New | 29.8B | Q6_K | 22.8 GB | 0.18 GB | 23.5 GB |
168.5 GB free
|
20 | MLX | Run |
| Granite 4.1 30B | 28.9B | Q6_K | 22.1 GB | 2.00 GB | 24.7 GB |
167.3 GB free
|
20 | MLX | Run |
| Qwen3.8 27B New this week | 27.8B | Q6_K | 21.2 GB | 0.50 GB | 22.3 GB |
169.7 GB free
|
21 | MLX | Run |
| Qwen3.6 27B | 27.8B | Q6_K | 21.2 GB | 0.50 GB | 22.3 GB |
169.7 GB free
|
21 | MLX | Run |
| Gemma 3 27B | 27.4B | Q6_K | 20.9 GB | 1.03 GB | 22.6 GB |
169.4 GB free
|
21 | MLX | Run |
| Gemma 4 26B-A4B | 26.5B MoE / 3.8B act | Q6_K | 20.2 GB | 0.51 GB | 21.3 GB |
170.7 GB free
|
72 | MLX | Run |
| Devstral Small 2 24B | 24B | Q6_K | 18.3 GB | 1.25 GB | 20.2 GB |
171.8 GB free
|
24 | MLX | Run |
| Mistral Small 3.2 24B | 23.6B | Q6_K | 18.0 GB | 1.25 GB | 19.9 GB |
172.1 GB free
|
25 | MLX | Run |
| gpt-oss 20B | 20.9B MoE / 3.6B act | MXFP4 | 10.8 GB | 0.19 GB | 11.6 GB |
180.4 GB free
|
112 | MLX | Run |
| Qwen3 14B | 14.8B | Q6_K | 11.3 GB | 1.25 GB | 13.2 GB |
178.8 GB free
|
40 | MLX | Run |
| DeepSeek-R1-Distill-Qwen 14B | 14.8B | Q6_K | 11.3 GB | 1.50 GB | 13.4 GB |
178.6 GB free
|
40 | MLX | Run |
| Qwen2.5-Coder 14B | 14.8B | Q6_K | 11.3 GB | 1.50 GB | 13.4 GB |
178.6 GB free
|
40 | MLX | Run |
| Phi-4 14B | 14.7B | Q6_K | 11.2 GB | 1.56 GB | 13.4 GB |
178.6 GB free
|
40 | MLX | Run |
| Ministral 3 14B | 13.9B | Q6_K | 10.6 GB | 1.25 GB | 12.5 GB |
179.5 GB free
|
42 | MLX | Run |
| Gemma 3 12B | 12.2B | Q6_K | 9.3 GB | 0.81 GB | 10.7 GB |
181.3 GB free
|
48 | MLX | Run |
| Mistral NeMo 12B | 12.2B | Q6_K | 9.3 GB | 1.25 GB | 11.2 GB |
180.8 GB free
|
48 | MLX | Run |
| Gemma 4 12B | 12B | Q6_K | 9.2 GB | 0.81 GB | 10.6 GB |
181.4 GB free
|
49 | MLX | Run |
| Qwen3.5 9B | 9.65B | Q6_K | 7.4 GB | 0.25 GB | 8.2 GB |
183.8 GB free
|
61 | MLX | Run |
| Ornith 1.5 9B New this week | 9.41B | Q6_K | 7.2 GB | 0.25 GB | 8.0 GB |
184.0 GB free
|
62 | MLX | Run |
| Ministral 3 8B | 8.92B | Q6_K | 6.8 GB | 1.06 GB | 8.5 GB |
183.5 GB free
|
66 | MLX | Run |
| Granite 4.1 8B | 8.79B | Q6_K | 6.7 GB | 1.25 GB | 8.6 GB |
183.4 GB free
|
67 | MLX | Run |
| LFM2.5 8B-A1B | 8.47B MoE / 1.5B act | Q6_K | 6.5 GB | 0.09 GB | 7.2 GB |
184.8 GB free
|
182 | MLX | Run |
| Fara 7B | 8.29B | Q6_K | 6.3 GB | 0.44 GB | 7.4 GB |
184.6 GB free
|
71 | MLX | Run |
| Qwen3 8B | 8.19B | Q6_K | 6.3 GB | 1.13 GB | 8.0 GB |
184.0 GB free
|
71 | MLX | Run |
| Llama 3.1 8B Instruct | 8.03B | Q6_K | 6.1 GB | 1.00 GB | 7.7 GB |
184.3 GB free
|
73 | MLX | Run |
| Gemma 4 E4B | 8.0B | Q6_K | 6.1 GB | 0.14 GB | 6.9 GB |
185.1 GB free
|
73 | MLX | Run |
| Ling 3.0 Tiny 7.9B-A1.3B New | 7.9B MoE / 1.3B act | Q6_K | 6.0 GB | 0.05 GB | 6.7 GB |
185.3 GB free
|
210 | MLX | Run |
| DeepSeek-R1-Distill-Qwen 7B | 7.62B | Q6_K | 5.8 GB | 0.44 GB | 6.9 GB |
185.1 GB free
|
77 | MLX | Run |
| Qwen2.5-Coder 7B | 7.62B | Q6_K | 5.8 GB | 0.44 GB | 6.9 GB |
185.1 GB free
|
77 | MLX | Run |
| Olmo 3 7B Instruct | 7.3B | Q6_K | 5.6 GB | 2.50 GB | 8.7 GB |
183.3 GB free
|
80 | MLX | Run |
| Mistral 7B Instruct v0.3 | 7.25B | Q6_K | 5.5 GB | 1.00 GB | 7.1 GB |
184.9 GB free
|
81 | MLX | Run |
| Gemma 4 E2B | 5.1B | Q6_K | 3.9 GB | 0.07 GB | 4.6 GB |
187.4 GB free
|
115 | MLX | Run |
| Qwen3.5 4B | 4.66B | Q6_K | 3.6 GB | 0.25 GB | 4.4 GB |
187.6 GB free
|
126 | MLX | Run |
| Gemma 3 4B | 4.3B | Q6_K | 3.3 GB | 0.30 GB | 4.2 GB |
187.8 GB free
|
136 | MLX | Run |
| Qwen3 4B | 4.02B | Q6_K | 3.1 GB | 1.13 GB | 4.8 GB |
187.2 GB free
|
146 | MLX | Run |
| Ministral 3 3B | 3.85B | Q6_K | 2.9 GB | 0.81 GB | 4.4 GB |
187.6 GB free
|
152 | MLX | Run |
| Phi-4-mini 3.8B | 3.84B | Q6_K | 2.9 GB | 1.00 GB | 4.5 GB |
187.5 GB free
|
152 | MLX | Run |
| Granite 4.1 3B | 3.4B | Q6_K | 2.6 GB | 0.63 GB | 3.8 GB |
188.2 GB free
|
172 | MLX | Run |
| Llama 3.2 3B Instruct | 3.21B | Q6_K | 2.5 GB | 0.88 GB | 3.9 GB |
188.1 GB free
|
182 | MLX | Run |
| LFM2.5 2.6B New | 2.7B | Q6_K | 2.1 GB | 0.13 GB | 2.8 GB |
189.2 GB free
|
217 | MLX | Run |
| Qwen3.5 2B | 2.27B | Q6_K | 1.7 GB | 0.09 GB | 2.4 GB |
189.6 GB free
|
258 | MLX | Run |
| Qwen3 1.7B | 1.72B | Q6_K | 1.3 GB | 0.88 GB | 2.8 GB |
189.2 GB free
|
340 | MLX | Run |
| Llama 3.2 1B Instruct | 1.24B | Q6_K | 0.9 GB | 0.25 GB | 1.8 GB |
190.2 GB free
|
472 | MLX | Run |
| Gemma 3 1B | 1.0B | Q6_K | 0.8 GB | 0.04 GB | 1.4 GB |
190.6 GB free
|
585 | MLX | Run |
| Qwen3.5 0.8B | 0.87B | Q6_K | 0.7 GB | 0.09 GB | 1.4 GB |
190.6 GB free
|
673 | MLX | Run |
| Qwen3 0.6B | 0.6B | Q6_K | 0.5 GB | 0.88 GB | 1.9 GB |
190.1 GB free
|
976 | MLX | Run |
Runs — tight
1 model · fits at 8K, drop context or quant to go longer| Model | Params | Quant | Weights | +KV 8K | Total | VRAM headroom | Tok/s | Runtime | |
|---|---|---|---|---|---|---|---|---|---|
| Qwen3 235B-A22B | 235B MoE / 22B act | Q6_K | 179.5 GB | 1.47 GB | 181.5 GB |
10.5 GB free
|
12 | MLX | Run |
Won't fit in VRAM
9 models · some are possible with CPU offload| Model | Params | Quant | Weights | +KV 8K | Total | Over budget | Tok/s | Runtime | |
|---|---|---|---|---|---|---|---|---|---|
| Kimi K3 2.8T-A104B New | 2780B MoE / 104B act | Q6_K | 2123.0 GB | 0.21 GB | 2123.9 GB | +1931.9 GB | ~2.6 | MLX | Why |
| Qwen3.8 2.4T-A95B New | 2446B MoE / 95B act | Q6_K | 1868.0 GB | 0.72 GB | 1869.3 GB | +1677.3 GB | ~2.9 | MLX | Why |
| DeepSeek V4 Pro 1.6T-A49B New | 1650B MoE / 49B act | Q6_K | 1260.1 GB | 0.54 GB | 1261.2 GB | +1069.2 GB | ~5.6 | MLX | Why |
| Kimi K2.6 1T-A32B | 1027B MoE / 32B act | Q6_K | 784.3 GB | 0.54 GB | 785.4 GB | +593.4 GB | ~8.5 | MLX | Why |
| GLM-5.2 744B-A40B | 753B MoE / 40B act | Q6_K | 575.1 GB | 0.69 GB | 576.3 GB | +384.3 GB | ~6.8 | MLX | Why |
| DeepSeek-R1 671B | 671B MoE / 37B act | Q6_K | 512.4 GB | 0.54 GB | 513.6 GB | +321.6 GB | ~7.4 | MLX | Why |
| Qwen3.5 397B-A17B | 403B MoE / 17B act | Q6_K | 307.8 GB | 0.23 GB | 308.6 GB | +116.6 GB | ~16 | MLX | Why |
| Ornith 1.5 397B-A17B New this week | 397B MoE / 17B act | Q6_K | 303.2 GB | 0.23 GB | 304.0 GB | +112.0 GB | ~16 | MLX | Why |
| DeepSeek V4 Flash 284B-A13B New | 284B MoE / 13B act | Q6_K | 216.9 GB | 0.38 GB | 217.9 GB | +25.9 GB | ~21 | MLX | Why |
Totals = quantised weights + f16 KV cache at 8K tokens + 0.6 GB runtime overhead, against 192.0 GB usable unified memory. Tokens per second are estimated from 800 GB/s peak memory bandwidth at batch 1 — not measured. Read the method.