Canonical Rankings

Best models for this Mac

Mac Studio M4 Max 64GB ranked for coding with a most capable bias.

Mac

Capability

Bias

26 ranked modelsCatalog current through February 27, 2026M5 Max watch: 8B 61.6 tok/s · 14B 34.3 tok/s on the current 36GB public anchor.

Rank	Model	Score	Quant	Tok/s	Evidence	Headroom	Why it ranks here
1	Qwen 3 32B	274	Q8_0	22.0 tok/s	Benchmark estimate	31.0 GB	Frontier candidate in the current catalog. Near-lossless quantization quality.
2	Gemma 3 27B	259	Q8_0	14.5 tok/s	Benchmark estimate	34.1 GB	Near-lossless quantization quality. Enough for most serious sessions.
3	DeepSeek R1 Distill Qwen 32B	252	Q8_0	—	Fit estimate	31.0 GB	Near-lossless quantization quality. Enough for most serious sessions.
4	Qwen3.5-27B	244	Q8_0	—	Fit estimate	36.4 GB	Frontier candidate in the current catalog. Near-lossless quantization quality.
5	Mistral Small 3.1 24B	240	Q8_0	—	Fit estimate	39.9 GB	Near-lossless quantization quality. Comfortable for long-context work.
6	Devstral Small 1.1	240	Q8_0	—	Fit estimate	39.9 GB	Near-lossless quantization quality. Comfortable for long-context work.
7	Devstral Small 2 24B	240	Q8_0	—	Fit estimate	39.9 GB	Frontier candidate in the current catalog. Near-lossless quantization quality.
8	Magistral Small	240	Q8_0	—	Fit estimate	39.9 GB	Near-lossless quantization quality. Enough for most serious sessions.
9	Llama 3.3 70B	233	Q5_K_M	—	Fit estimate	14.2 GB	High quality quantization quality. Fine for standard chats and coding.
10	Qwen 3 30B-A3B	233	Q8_0	84.9 tok/s	Benchmark estimate	34.3 GB	Near-lossless quantization quality. Comfortable for long-context work.
11	Qwen 3 8B	211	Q8_0	63.1 tok/s	Benchmark estimate	54.7 GB	Near-lossless quantization quality. Comfortable for long-context work.
12	Qwen3-Coder-30B-A3B	211	Q8_0	—	Fit estimate	34.7 GB	Near-lossless quantization quality. Comfortable for long-context work.
13	Qwen3.5-35B-A3B	210	Q8_0	—	Fit estimate	30.3 GB	Near-lossless quantization quality. Comfortable for long-context work.
14	GLM-4.7-Flash	210	Q8_0	—	Fit estimate	28.2 GB	Near-lossless quantization quality. Fine for standard chats and coding.
15	Llama 2 7B	209	Q8_0	36.4 tok/s	Benchmark estimate	53.2 GB	Near-lossless quantization quality. Context headroom is tight.
16	Qwen 2.5 14B	199	Q8_0	—	Fit estimate	49.0 GB	Near-lossless quantization quality. Comfortable for long-context work.
17	Phi-4 14B	198	Q8_0	—	Fit estimate	49.5 GB	Near-lossless quantization quality. Fine for standard chats and coding.
18	Qwen3.5-9B	191	Q8_0	—	Fit estimate	54.1 GB	Frontier candidate in the current catalog. Near-lossless quantization quality.
19	Llama 3.1 8B	189	Q8_0	—	Fit estimate	55.0 GB	Near-lossless quantization quality. Comfortable for long-context work.
20	Qwen 2.5 7B	189	Q8_0	—	Fit estimate	56.0 GB	Near-lossless quantization quality. Comfortable for long-context work.
21	Mistral 7B v0.3	188	Q8_0	—	Fit estimate	55.7 GB	Near-lossless quantization quality. Enough for most serious sessions.
22	Gemma 3 4B	150	Q8_0	100.5 tok/s	Benchmark estimate	58.4 GB	Near-lossless quantization quality. Comfortable for long-context work.
23	Qwen 3 4B	150	Q8_0	143.2 tok/s	Benchmark estimate	58.6 GB	Near-lossless quantization quality. Enough for most serious sessions.
24	Qwen 3 0.6B	145	Q8_0	184.4 tok/s	Benchmark estimate	62.8 GB	Near-lossless quantization quality. Enough for most serious sessions.
25	Qwen3-Coder-Next	134	Q5_K_M	—	Fit estimate	9.8 GB	High quality quantization quality. Fine for standard chats and coding.
26	Llama 3.2 1B	124	Q8_0	—	Fit estimate	62.1 GB	Near-lossless quantization quality. Comfortable for long-context work.