Check my GPU

on

Can the Apple M2 Pro (32 GB) run ERNIE 4.5 21B-A3B Thinking?

84 ptsGreat (A) · 84 of 100 points: Model quality 29.9/45, Memory fit 25.0/25, Speed 14.1/15, Memory use 15.0/15

Yes, it runs well

Runs well. At Q4_K_M it needs about 16.1 GB of 21.4 GB usable, with an estimated speed of 35.2 tok/s.

Best quant
Q4_K_M
Generation speed
35.2 tok/s
Prompt processing
1,405 tok/s
First token (1K prompt)
0.7 s
Largest safe context
98K

Download ERNIE 4.5 21B-A3B Thinking

Recommended for the Apple M2 Pro (32 GB): Q4_K_M. Every other quantization has its own download link in the table below.

Quantized files · GGUF

Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.

Download the Q4_K_M file (opens in a new tab)

huggingface.co/unsloth/ERNIE-4.5-21B-A3B-Thinking-GGUF (opens in a new tab)

ollama run hf.co/unsloth/ERNIE-4.5-21B-A3B-Thinking-GGUF:Q4_K_M

Original weights · Hugging Face

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

huggingface.co/baidu/ERNIE-4.5-21B-A3B-Thinking (opens in a new tab)

No sign-in needed.

Official repo · GitHub

Reference code, inference scripts and docs from Baidu. The weights themselves download from Hugging Face.

github.com/PaddlePaddle/ERNIE (opens in a new tab)

git clone https://github.com/PaddlePaddle/ERNIE

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

How the A 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.29.9 / 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.14.1 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.15.0 / 15
Total84.0 / 10084 ptsGreat (A) · 84 of 100 points: Model quality 29.9/45, Memory fit 25.0/25, Speed 14.1/15, Memory use 15.0/15

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 11.5 GB 21.4 GB 57.9 tok/s Runs well 74 ptsGreat (A) · 74 of 100 points: Model quality 22.1/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 12.4/15 GGUF (opens in a new tab)
Q3_K_M 13.5 GB 21.4 GB 45.6 tok/s Runs well 80 ptsGreat (A) · 80 of 100 points: Model quality 26.5/45, Memory fit 25.0/25, Speed 15.0/15, Memory use 14.5/15 GGUF (opens in a new tab)
Q4_K_M 16.1 GB 21.4 GB 35.2 tok/s Runs well 84 ptsGreat (A) · 84 of 100 points: Model quality 29.9/45, Memory fit 25.0/25, Speed 14.1/15, Memory use 15.0/15 GGUF (opens in a new tab)
Q5_K_M 18.3 GB 21.4 GB 29.1 tok/s Tight fit 82 ptsGreat (A) · 82 of 100 points: Model quality 30.6/45, Memory fit 24.0/25, Speed 12.8/15, Memory use 15.0/15 GGUF (opens in a new tab)
Q6_K 20.7 GB 21.4 GB 24.4 tok/s Tight fit 81 ptsGreat (A) · 81 of 100 points: Model quality 31.2/45, Memory fit 24.0/25, Speed 11.7/15, Memory use 15.0/15 GGUF (opens in a new tab)
Q8_0 25.9 GB 21.4 GB 15.2 tok/s Heavy offload 59 ptsGood (B) · 59 of 100 points: Model quality 31.5/45, Memory fit 4.0/25, Speed 9.3/15, Memory use 15.0/15 GGUF (opens in a new tab)
F16 46.4 GB 21.4 GB — Won't run 9 ptsWon't run (F) · 9 of 100 points: Model quality 9.5/45, Memory fit 0.0/25, Speed 0.0/15, Memory use 0.0/15 GGUF (opens in a new tab) 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

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.