Every open model in our directory, ranked by how well it runs on Apple M4 Pro (14-core CPU, 20-core GPU) (64GB) at 4-bit. Grades weigh whether the model fits in memory and how fast it should run.
Apple M4 Pro (14-core CPU, 20-core GPU) has 64 GB of VRAM, enough to run 135 of the 169 open models we track at 4-bit. The largest that still fits well is Kimi K2 Instruct (1000B), and the top-rated is Qwen3-30B-A3B at about 41 tok/s.
135 of 169 models run comfortably on Apple M4 Pro (14-core CPU, 20-core GPU).
See full Apple M4 Pro (14-core CPU, 20-core GPU) specs and pricingThe top models this device can run at 4-bit, ranked by fit and speed.
| Model | Grade | Speed | VRAM |
|---|---|---|---|
| Qwen3-30B-A3BAlibaba | SS | 40.8 tok/s | 5.4 GB |
| Llama 3 8B InstructMeta | AA | 38.8 tok/s | 5.7 GB |
| LFM2.5-8B-A1BLiquid AI | AA | 75.6 tok/s | 2.9 GB |
| Carnice-9b for Hermes agentkai-os | AA | 36.5 tok/s | 6.0 GB |
| PersonaPlex 7B |
Qwen3-30B-A3B is the highest-rated model that runs on Apple M4 Pro (14-core CPU, 20-core GPU) at 4-bit, at about 41 tok/s. In total, Apple M4 Pro (14-core CPU, 20-core GPU) can run 135 of the 169 open models we track.
We size each model at 4-bit (Q4_K_M), add the KV cache for its context length and a fixed runtime overhead, then compare that to this device VRAM. Each model is graded on how comfortably it fits and how fast it should run on this card.
It is an estimate of decode speed in tokens per second at 4-bit, based on this device memory bandwidth and each model size. Real speed depends on your runtime, batch size, and context length, so use it to compare models, not as a hard promise.
Lower-bit quantization shrinks a model so more of them fit. These grades already assume 4-bit, which is the common local default. Going below 4-bit can squeeze a larger model in at some quality cost; going above needs more VRAM and may not fit.
| AA |
| 45.9 tok/s |
| 4.8 GB |
| Llama 2 7B ChatMeta | AA | 45.9 tok/s | 4.8 GB |
| North Mini CodeCohere | AA | 26.2 tok/s | 8.4 GB |
| Nemotron 3 Nano OmniNVIDIA | AA | 25.8 tok/s | 8.5 GB |
| Qwen3.6 35B-A3BAlibaba | AA | 25.8 tok/s | 8.5 GB |
| Qwen3.5-35B-A3BAlibaba | AA | 25.8 tok/s | 8.5 GB |
| Gemma 4 E2B ITGoogle | AA | 59.3 tok/s | 3.7 GB |
| VibeThinker-3BWeiboAI | AA | 57.6 tok/s | 3.8 GB |
Qwen3-30B-A3BAlibaba | 30B(3B active) | SS | 40.8 tok/s | 5.4 GB | |
| 8B | AA | 38.8 tok/s | 5.7 GB | ||
LFM2.5-8B-A1BLiquid AI | 8.3B(1.5B active) | AA | 75.6 tok/s | 2.9 GB | |
| 9B | AA | 36.5 tok/s | 6.0 GB | ||
PersonaPlex 7BNVIDIA | 7B | AA | 45.9 tok/s | 4.8 GB | |
Llama 2 7B ChatMeta | 7B | AA | 45.9 tok/s | 4.8 GB | |
North Mini CodeCohere | 30B(3B active) | AA | 26.2 tok/s | 8.4 GB | |
Nemotron 3 Nano OmniNVIDIA | 30B(3B active) | AA | 25.8 tok/s | 8.5 GB | |
Ad CodeRabbitAI-powered Code Reviews. Cut review time & bugs in half, instantly.Try for Free | |||||
Qwen3.6 35B-A3BAlibaba | 35B(3B active) | AA | 25.8 tok/s | 8.5 GB | |
Qwen3.5-35B-A3BAlibaba | 35B(3B active) | AA | 25.8 tok/s | 8.5 GB | |
Gemma 4 E2B ITGoogle | 2B | AA | 59.3 tok/s | 3.7 GB | |
VibeThinker-3BWeiboAI | 3B | AA | 57.6 tok/s | 3.8 GB | |
Mistral 7B InstructMistral AI | 7B | AA | 34.4 tok/s | 6.4 GB | |
Llama 2 13B ChatMeta | 13B | AA | 26.0 tok/s | 8.5 GB | |
Gemma 4 E4B ITGoogle | 4B | AA | 31.8 tok/s | 6.9 GB | |
Gemma 3 4B ITGoogle | 4B | AA | 31.8 tok/s | 6.9 GB | |
Ad PayPerQPay-per-query access to top LLMs without a subscription. Use any model on demand.Try PayPerQ | |||||
Mixtral 8x7B InstructMistral AI | 46.7B(12.9B active) | AA | 19.3 tok/s | 11.4 GB | |
DiffusionGemma 26B-A4BGoogle | 25.2B(3.8B active) | AA | 21.0 tok/s | 10.5 GB | |
Gemma 4 26B-A4B ITGoogle | 26B(4B active) | AA | 20.0 tok/s | 11.0 GB | |
Qwen3.5-122B-A10BAlibaba | 122B(10B active) | BB | 8.1 tok/s | 27.3 GB | |
Qwen3-235B-A22BAlibaba | 235B(22B active) | BB | 6.0 tok/s | 36.3 GB | |
minimax-m2.5MiniMax | 230B(10B active) | BB | 9.7 tok/s | 22.7 GB | |
Llama 2 70B ChatMeta | 70B | BB | 5.1 tok/s | 43.4 GB | |
Mixtral 8x22B InstructMistral AI | 141B(39B active) | BB | 5.0 tok/s | 43.6 GB | |
Ad Vast.aiAffordable on-demand GPU rentals for training and inference. Pick from thousands of hosts.Rent a GPU | |||||
Qwen 3.5 OmniAlibaba | 397B(17B active) | BB | 4.9 tok/s | 45.2 GB | |
Model Compatibility
Use the calculator to weigh this device against any other, model by model, with speed and fit scores.
Mixture-of-experts models only activate a fraction of their parameters at a time, so their runtime memory is closer to the active size than the headline size. We score fit on the active parameters, which is why some very large models still fit.
