CogVLM2 Llama 3 Chat 19B on Apple M3 Ultra (96 GB)

Can the Apple M3 Ultra (96 GB) run CogVLM2 Llama 3 Chat 19B?

Yes, it runs great

Runs great. At Q4_K_M it needs about 22.5 GB of 72.0 GB usable, with an estimated speed of 44.4 tok/s.

Best quant: Q4_K_M
Generation speed: 44.4 tok/s
Prompt processing: 903 tok/s
First token (1K prompt): 1.1 s
Largest safe context: 8K

Download CogVLM2 Llama 3 Chat 19B

Original weights · Hugging Face

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

huggingface.co/zai-org/cogvlm2-llama3-chat-19B (opens in a new tab)

No sign-in needed.

Links checked 2 days ago against Hugging Face, Ollama and GitHub.

How it’s rated

Quality 36Basic quality (D) · 36 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at Q4_K_M, from its editorial quality tier scaled by the quantization or precision. It’s the same on any hardware that runs this variant. Scale: A+ 85+ · A 70+ · B 55+ · C 40+ · D under 40.
Runs Runs great: How well it runs on the Apple M3 Ultra (96 GB): speed, capped by how well it fits in memory. Runs great: 30+ tokens/s (or near the image, video or decision speed target) · Runs well: 10+ tokens/s · Slow: 3+ tokens/s · Very slow: under 3 · Won’t run: doesn’t fit even with offload. A tight fit or RAM offload caps it at Runs well; heavy offload caps it at Slow.

Match score

To order lists, each result also gets a match score out of 100 from four parts. It favors the most capable model that still runs well.

Model qualityEditorial quality tier of the model, scaled by the quantization or precision used.	16.2 / 45
Memory fitFull marks when the model fits in the share of unified memory the GPU can use; fewer when it goes past that limit.	25.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	15.0 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.	7.2 / 15
Match score	63.5 / 100

The match score orders results; it is not shown as a grade. Quality and Runs are the two ratings in the tables.

Every quantization on the Apple M3 Ultra (96 GB)

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	18.5 GB	72.0 GB	72.9 tok/s	Runs great	26Basic quality (D) · 26 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 57/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q3_K_M`	20.2 GB	72.0 GB	57.4 tok/s	Runs great	31Basic quality (D) · 31 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 60/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q4_K_M`	22.5 GB	72.0 GB	44.4 tok/s	Runs great	36Basic quality (D) · 36 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q5_K_M`	24.5 GB	72.0 GB	36.7 tok/s	Runs great	36Basic quality (D) · 36 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q6_K`	26.6 GB	72.0 GB	30.8 tok/s	Runs great	37Basic quality (D) · 37 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`Q8_0`	31.3 GB	72.0 GB	22.6 tok/s	Runs well	38Basic quality (D) · 38 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking.	All files (opens in a new tab)
`F16`	49.7 GB	72.0 GB	10.4 tok/s	Runs well	38Basic quality (D) · 38 of 100: how capable the model is at F16, whatever it runs on. Match score 64/100 (quality, fit, speed and memory use) sets the ranking.	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

38Basic quality (D) · 38 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 72/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 5090 Q8_0 · 57.4 tok/s · Runs great
36Basic quality (D) · 36 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. AMD Radeon RX 7900 XTX Q5_K_M · 43.0 tok/s · Runs great
36Basic quality (D) · 36 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 3090 Q5_K_M · 48.7 tok/s · Runs great
36Basic quality (D) · 36 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 71/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4090 Q5_K_M · 52.4 tok/s · Runs great
31Basic quality (D) · 31 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking. AMD Radeon RX 9070 XT Q3_K_M · 44.9 tok/s · Runs well (tight fit)
31Basic quality (D) · 31 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4070 Ti SUPER Q3_K_M · 54.7 tok/s · Runs well (tight fit)

Quick answers

How much memory does CogVLM2 Llama 3 Chat 19B need on the M3 Ultra 96GB?

At Q4_K_M it needs about 22.5 GB. The M3 Ultra 96GB has 72.0 GB usable.

How fast is CogVLM2 Llama 3 Chat 19B on the M3 Ultra 96GB?

About 44.4 tok/s at Q4_K_M, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of CogVLM2 Llama 3 Chat 19B should I use on the M3 Ultra 96GB?

Q4_K_M. Of the versions that fit, it is the one that runs best; when two run equally well, the check prefers the higher-rated one.

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.