MiniCPM-V 4.7 35B-A3B runs at 48 tok/s (48.2 as recorded) on a 64 GB M4 Pro, a community run reported by omlx.ai (MLX, 5-bit, 1k context). It needs about 23 GB of unified memory. It is OpenBMB's 35.2B mixture-of-experts model (3B active), released under the Apache 2.0 licence.
Want to run MiniCPM-V 4.7 35B-A3B faster — or step up to bigger models? Find your Mac →This is a community run: a real run, reported by its source and linked below. The index records it as given and never adjusts it.
| Metric | Value |
|---|---|
| Tokens per second | community48 tok/s |
| Provenance | community run · omlx.ai |
| Quantization | 5-bit |
| Engine | MLX |
| Context | 1k |
| Figure dated | 2026-10-16 |
| Minimum RAM | 23 GB |
| Parameters | 35.2B MoE (3B active) |
| Chip bandwidth | 273 GB/s |
Each row as the index holds it, with its source. “Mac RAM” is the memory of the Mac a run used, not the model’s requirement (23 GB). The headline figure follows the reference rule: the lower-median measured plain-decode run at 5 bits or fewer (a run whose quant was not stated counts only when no stated-quant run qualifies), else the estimate; a run marked “not used to anchor estimates” is listed here but never becomes the headline.
| Speed | Quant | Engine | Mac RAM | Context | Source | Note |
|---|---|---|---|---|---|---|
| community48.2 tok/s | 5-bit | MLX | 64 GB | 1k | omlx.ai | 16-core GPU. oMLX v0.7.1.dev2, MTP/DFlash off. 47.9 tok/s at 4k. |
Ollama — install it (brew install ollama or the app from ollama.com), then pull the model by its tag, checked on ollama.com on 2026-10-17:
MLX — pick a build from Hugging Face’s MLX listings, then run it with its repo id:
Downloads: Hugging Face repo → · model card →
The 48 tok/s figure above came from MLX at 5-bit; other runtimes and quants run at different speeds.
Prefer a GUI? LM Studio downloads GGUF and MLX builds in one click. Step by step: install Ollama on a Mac.
| Chip | Speed | Quant | Engine | Bandwidth |
|---|---|---|---|---|
| M5 Max | community100 tok/s | MXFP4 | MLX | 614 GB/s |
| M4 Max | estimated92 tok/s | Q4_K_M | MLX | 546 GB/s |
| M5 Pro | estimated59 tok/s | Q4_K_M | MLX | 307 GB/s |
| M6 | estimated37 tok/s | Q4_K_M | MLX | 170 GB/s |
OpenBMB's Apache 2.0 vision-language MoE (Oct 2026): 35B total with ~3B active, image, multi-image and video input, 256K context. Q4_K_M is 21 GB plus a 1.1 GB vision projector, so it fits a 24 GB Mac; it runs in llama.cpp b11551+ (an image-grounding fix, PR #30510, is still open), mlx-vlm 0.7.8 and OpenBMB's Ollama namespace. Community oMLX runs measured 100.2 tok/s on an M5 Max (MXFP4) and 48.2 on an M4 Pro (5-bit).
Apache 2.0 licence · 35.2B MoE · 262K context · LLMCheck capability 35/50 · see its leaderboard rank →
MiniCPM-V 4.7 35B-A3B needs about 23 GB of unified memory. These current Apple Silicon Macs have the headroom to run it comfortably:
As an Amazon Associate, LLMCheck earns from qualifying purchases. These affiliate links cost you nothing extra and help keep the LLMCheck index free.
See how MiniCPM-V 4.7 35B-A3B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard