Newshot

ModelBest: MiniCPM Family Models Surpass 70 Million Global Downloads as of November 2026

AI summary

ModelBest announced today that global downloads of its MiniCPM family models have exceeded 70 million as of November 2026. ModelBest was founded in Beijing in August 2022, incubated from Tsinghua University's Natural Language Processing Lab. In October this year, ModelBest announced the launch of its next-generation on-device multimodal large model MiniCPM-V 4.7, offered in 35B-A3B and 1B variants, with 35B-A3B being the first MoE model in the MiniCPM-V series. In July this year, Samsung officially launched its new foldable flagships Galaxy Z Fold8, Galaxy Z Fold8 Ultra, and Galaxy Z Flip8, and ModelBest announced that its MiniCPM series on-device large models deeply empower Samsung Galaxy AI—the first time a Chinese on-device large model has entered the global flagship product line of a leading international smartphone maker.

Translation(model translation, not the source's own text)

ModelBest: MiniCPM Family Models Surpass 70 Million Global Downloads as of November 2026

IT Home, November 19 — ModelBest officially announced today that, as of November 2026, the MiniCPM family of models has surpassed 70 million global downloads.

IT Home note: ModelBest was founded in Beijing in August 2022, incubated from the Natural Language Processing Laboratory of Tsinghua University. Co-founder and CEO 李大海 previously served as a partner and CTO of Zhihu; co-founder and Chief Scientist 刘知远 is currently a professor in the Department of Computer Science and Technology at Tsinghua University; co-founder and CTO 曾国洋, at the age of 22, served as the core engineering lead and trained China's first large language model, CPM-1.

In October this year, ModelBest announced the launch of its new-generation on-device multimodal large model MiniCPM-V 4.7, offering two specifications: 35B-A3B and 1B, with 35B-A3B being the first MoE model in the MiniCPM-V series.

According to the official introduction, this model is based on the latest LES (Lookahead Expert Scheduling) inference acceleration technology. The 35B-A3B version is the first open-source MoE model to support lossless, smooth on-device operation, achieving a decoding speed of 13+ token/s on the iPhone 17 Pro Max with only 3GB of memory.

In July this year, Samsung's all-new foldable flagships — Galaxy Z Fold8, Galaxy Z Fold8 Ultra, and Galaxy Z Flip8 — were officially released. ModelBest announced that its MiniCPM series on-device large models deeply empower Samsung Galaxy AI. This marks the first time a domestic on-device large model has entered the global flagship product line of a leading international smartphone manufacturer.

Related Articles

  • ModelBest Releases New-Generation On-Device Multimodal Large Model MiniCPM-V 4.7, Achieving 13+ token/s Decoding Speed with Only 3GB of Memory
  • ModelBest Open-Sources MiniCPM5-2B AI Model: Ranked No. 1 Globally Under 4B on AA Leaderboard, Beginning to Show On-Device General Agent Capabilities
  • ModelBest: MiniCPM Series Open-Source Models Surpass 50 Million Global Cumulative Downloads
  • AI Company ModelBest Initiates IPO
  • First for a Domestic On-Device Large Model: ModelBest's MiniCPM Officially Announced for Samsung Galaxy Z Fold8 Series Phones
  • ModelBest Releases On-Device Model MiniCPM5-2B, Ranked No. 1 in Performance Under 4B on AA-Index Leaderboard

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by IT之家
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗