LLaVA 1.6 Mistral 7B on Apple M2 Pro (32 GB)

Can the Apple M2 Pro (32 GB) run LLaVA 1.6 Mistral 7B?

57 pts

Yes, it runs well

Runs well. At Q3_K_M it needs about 7.1 GB of 21.4 GB usable, with an estimated speed of 36.1 tok/s.

Best quant: Q3_K_M
Generation speed: 36.1 tok/s
Prompt processing: 557 tok/s
First token (1K prompt): 1.8 s
Largest safe context: 32K

Download LLaVA 1.6 Mistral 7B

Recommended for the Apple M2 Pro (32 GB): Q3_K_M. Every other quantization has its own download link in the table below.

Quickest · Ollama

One command downloads a ready-to-run version and starts a chat.

ollama run llava:7b-v1.6-mistral-q3_K_M

Get Ollama at ollama.com/download (opens in a new tab) · model page (opens in a new tab)

Original weights · Hugging Face

Full-precision files from LLaVA, for fine-tuning, vLLM, Transformers.

huggingface.co/llava-hf/llava-v1.6-mistral-7b-hf (opens in a new tab)

No sign-in needed.

Links checked 7 days ago against Hugging Face, Ollama and GitHub.

How the C grade is scored

Every result gets up to 100 points from four parts. The points decide the letter grade.

Model qualityEditorial quality tier of the model, scaled by the quantization or precision used.	10.6 / 45
Memory fitFull marks when everything sits in GPU or unified memory; fewer when part spills to system RAM.	25.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	14.3 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.	7.6 / 15
Total	57.5 / 100	57 ptsUsable (C) · 57 of 100 points: Model quality 10.6/45, Memory fit 25.0/25, Speed 14.3/15, Memory use 7.6/15

Grade scale: S 85+ · A 72+ · B 58+ · C 42+ · D 25+ · F under 25 points.

Every quantization on the Apple M2 Pro (32 GB)

Quant	Memory needed	Available	Speed	Fit	Grade · points	Download
`Q2_K`	6.4 GB	21.4 GB	45.9 tok/s	Runs well	55 ptsUsable (C) · 55 of 100 points: Model quality 8.8/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 6.9/15
`Q3_K_M`	7.1 GB	21.4 GB	36.1 tok/s	Runs well	57 ptsUsable (C) · 57 of 100 points: Model quality 10.6/45, Memory fit 25.0/25, Speed 14.3/15, Memory use 7.6/15
`Q4_K_M`	8.0 GB	21.4 GB	27.9 tok/s	Runs well	58 ptsGood (B) · 58 of 100 points: Model quality 12.0/45, Memory fit 25.0/25, Speed 12.5/15, Memory use 8.6/15
`Q5_K_M`	8.8 GB	21.4 GB	23.1 tok/s	Runs well	58 ptsGood (B) · 58 of 100 points: Model quality 12.2/45, Memory fit 25.0/25, Speed 11.4/15, Memory use 9.4/15
`Q6_K`	9.6 GB	21.4 GB	19.4 tok/s	Runs well	58 ptsGood (B) · 58 of 100 points: Model quality 12.5/45, Memory fit 25.0/25, Speed 10.4/15, Memory use 10.3/15
`Q8_0`	11.4 GB	21.4 GB	14.2 tok/s	Runs well	58 ptsGood (B) · 58 of 100 points: Model quality 12.6/45, Memory fit 25.0/25, Speed 8.9/15, Memory use 12.3/15
`F16`	18.5 GB	21.4 GB	6.6 tok/s	Tight fit	57 ptsUsable (C) · 57 of 100 points: Model quality 12.6/45, Memory fit 24.0/25, Speed 6.1/15, Memory use 15.0/15	Original weights (opens in a new tab)

Memory figures assume 8K of context. Longer conversations or documents grow the KV cache.

Other hardware that runs it

67 ptsGood (B) · 67 of 100 points: Model quality 12.6/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 AMD Radeon RX 9070 XT Q8_0 · 45.5 tok/s
67 ptsGood (B) · 67 of 100 points: Model quality 12.6/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 NVIDIA GeForce RTX 4070 Ti SUPER Q8_0 · 55.5 tok/s
67 ptsGood (B) · 67 of 100 points: Model quality 12.6/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 NVIDIA GeForce RTX 4080 SUPER Q8_0 · 60.8 tok/s
67 ptsGood (B) · 67 of 100 points: Model quality 12.6/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 NVIDIA GeForce RTX 5070 Ti Q8_0 · 74.0 tok/s
67 ptsGood (B) · 67 of 100 points: Model quality 12.6/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 NVIDIA GeForce RTX 5080 Q8_0 · 79.2 tok/s
67 ptsGood (B) · 67 of 100 points: Model quality 12.5/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 15.0/15 NVIDIA GeForce RTX 3060 12GB Q6_K · 40.5 tok/s

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.