Unverified50% confidenceFactExact time
Engineering practices publicly disclosed by OpenBMB enable the 35B-A3B MoE model MiniCPM-V 4.7 to run on an iPhone at 20.5 tok/s without skipping any experts.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/21/2026
First Seen
Valid until: 1/19/2027
Sources
iPhone Runs 35B MoE at 20.5 tok/s: The Scheduling Magic That Skips No Experts
redditr/LocalLLaMA10/21/2026
Related Entities
Related Claims
UnverifiedOpenBMB's MiniCPM-V 4.7 can run a 35B vision MoE on an iPhone with about 3GB.77% similarUnverifiedMiniCPM-V 4.7 35B-A3B achieves a decoding speed of 13+ tokens/s on the iPhone 17 Pro Max.67% similarUnverifiedMiniCPM和Qwen-0.5B等1B以下轻量模型配合INT4量化可以在手机芯片上运行67% similarUnverifiedOn an M4 24GB device, a 9B model (such as qwen3.5-32k:9b-mlx) runs fast with low memory usage, but is unable to effectively identify code issues.66% similarUnverifiedThe UP host tested MiniCPM-2B on a computer with 4G VRAM and achieved a generation speed of 57 tokens per second, with a first-token latency of only 0.22 seconds.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1081657API
curl https://kongchang.com/api/v1/knowledge/claims/1081657MCP
get_claim(id=1081657)