Developer docs
API playgroundTry for free, no card

Search company profiles

ModelBest

Full company profile

uuid0003f3k

Namestring
ModelBest
Legal namestring
北京面壁智能科技有限责任公司
Websiteurl
modelbest.cn
Company typeenum
Private
Founded yearint
2022
Descriptiontext

ModelBest (北京面壁智能科技有限责任公司) is a Beijing-based AI foundation-model company founded in 2022 that develops lightweight, high-performance large language and multimodal models optimized for on-device edge deployment. Its core product family is the MiniCPM series, which spans text LLMs (0.5B to 8B parameters), multimodal vision-language models (MiniCPM-V 2.0 through 4.5), the full-modal MiniCPM-o 2.6, and the VoxCPM voice generation model, all built on a proprietary theoretical framework the company calls the 'Knowledge Density Law' (面壁定律). The technology stack emphasizes sparse attention, int4 quantization, the CPM.cu inference framework, and chip-level co-optimization, enabling models to run on consumer-grade hardware with as little as 2GB of memory at up to 220x inference acceleration. The company open-sources its models through the OpenBMB community (100,000+ GitHub stars, 1M+ downloads) while commercializing through enterprise B2B licensing, OEM/chip integration, and bespoke solutions.

ModelBest monetizes primarily through model licensing and OEM integration fees (MiniCPM embedded in partners' products and platforms such as MediaTek mobile SoCs, Intel AI PC chips, and Baidu Cloud) and through professional services for customized industry solutions, including intelligent cabins, on-device AI assistants, and vertical-domain models (e.g., China's first judicial foundation model co-developed with the Supreme People's Court). Its go-to-market blends a product-led/open-source top-of-funnel with enterprise direct sales and chip-OEM channel partnerships targeting automotive OEMs (Geely, Changan, Great Wall Motors, Volkswagen, Huawei), consumer-electronics manufacturers, cloud platforms, robotics companies (Booster Robotics, Elephant Robot), and government/institutional buyers. The company has raised eight funding rounds between 2022 and April 2026, completed multiple Tier-1 automotive and chip OEM partnerships through 2025, and became the first LLM company in China to pass ASPICE L2 automotive-grade certification.

Short descriptiontext

ModelBest (面壁智能) is a Beijing-based AI company, founded in 2022, that develops lightweight on-device large language, multimodal, and voice models (MiniCPM, MiniCPM-V, MiniCPM-o, VoxCPM) and licenses them to automotive OEMs, chip makers, consumer-electronics manufacturers, robotics firms, and government institutions including China's Supreme People's Court.

Operating statusenum
Operating
Ownership categoryenum
Headcount rangeband
51–100
akta.pro rankint
HeadquartersBeijing, China
HQ citystring
Beijing
HQ countrystring
China
HQ regionstring
Asia
Markets served

Serves global market

Offices1 record

Each record includes

City, Country, Type, Description, Source

Keyword5 values
edge AI models, lightweight large language models, on-device AI inference, multimodal foundation models, AI model licensing
Industry5 codes
PRIMARYHDAAACAN
On-Device/Edge Foundation Models (Mobile/Embedded LLMs)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Foundation Models & Generative AI (LLMs, multimodal)
SECONDARYHDAEANAC
Model Serving, Inference & Deployment Platforms (APIs, Edge/On-Prem)
Hardware, Technology & Telecom/Enterprise Platforms & Databases/Artificial Intelligence Platforms for Enterprise
SECONDARYHDAAACAB
Model Hosting, Serving & Inference Platforms
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Foundation Models & Generative AI (LLMs, multimodal)
SECONDARYHDAAACAC
Fine-Tuning, Adaptation & Custom Model Training (PEFT/LoRA/RLHF)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Foundation Models & Generative AI (LLMs, multimodal)
SECONDARYBPAEAHAH
MLOps/LLMOps & Model Lifecycle Management Services
NAICS code3 codes
  • Software Publishers5132
  • Custom Computer Programming Services541511
  • Computer Systems Design and Related Services5415
SIC code3 codes
  • Services-Prepackaged Software7372
  • Services-Computer Programming Services7371
  • Services-Computer Integrated Systems Design7373
Product category
Edge AI Foundation Models
GTM motion1 record

Each record includes

Type, Description, Source

Revenue model2 records
1Model Licensing and OEM Integration
TypeLicensing Royalties
Description

License MiniCPM model series to enterprise customers including automotive OEMs, smartphone/PC manufacturers, and robotics companies. Models are integrated into partners' chips (MediaTek SoC), platforms (Intel AI PC, Baidu Cloud), and products. Revenue derived from technology licensing fees and per-unit royalties where applicable.

modelbest.cn
2Enterprise B2B Solutions
TypeProfessional Services
Description

Provide customized AI solutions for specific industries such as judicial/legal systems (first judicial judgment vertical domain model in China), automotive (intelligent cabins), and consumer electronics. Revenue from solution development fees and ongoing support contracts.

modelbest.cn
Marketing channels6 records

Each record includes

Title, Type, Stage, Description, Source

Distribution channels5 records

Each record includes

Title, Type, Scope, Target buyer, Description, Source

Cost components5 values
Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations
GTM typeB2B
B2B
Offering typeSoftware
Software
Brand1 of 5 records shown
1MiniCPM (小钢炮)
Description

World-leading lightweight and high-performance LLM series for edge devices including phones, PCs, AI cabins, and embodied robots.

modelbest.cn
+4 more records
Core offering1 text field

ModelBest develops and licenses the MiniCPM series of lightweight, high-performance large language and multimodal AI foundation models optimized for on-device (edge) deployment. The portfolio includes text LLMs (MiniCPM 4.0/3.0/2.0/1.0), vision-language models (MiniCPM-V 4.5/2.6/2.5/2.0), full-modal models (MiniCPM-o 2.6), and voice generation (VoxCPM). Models are delivered via open-source community distribution, enterprise licensing to OEMs and chip makers, and integration into consumer electronics, automotive intelligent cabins, embodied robots and government systems.

Differentiator
Functional benefit
Problem solved
Quantifiable outcome1 of 7 values shown
  • MiniCPM 4.0 achieves performance on par with Qwen-3-8B and surpassing Gemma-3-12B with only 22% training cost
+6 more records
Product overview1 text field

ModelBest is an edge AI company offering a comprehensive portfolio of efficient on-device large language models. The product suite centers on the MiniCPM series for text generation, MiniCPM-V for multimodal (image/video) understanding, MiniCPM-o for full-modal (audio/video/text/image) capabilities, and VoxCPM for voice generation. These models are deployed across consumer electronics (phones, PCs) via Pinea brand, automotive intelligent cabins, embodied robots, and wearable devices. The company pioneered the 'Knowledge Density Law' guiding efficient model development where parameter counts halve every 3.3 months for equivalent performance. Models support major inference frameworks (llama.cpp, ollama, vLLM) and are available through the OpenBMB open-source community.

Product and service2 records
1MiniCPM Series
CategoryOn-device Large Language Models
Description

Lightweight, high-performance on-device large language model family in 8B, 4B, 2.4B, 1.2B and 0.5B variants that delivers ChatGPT- to GPT-4o-level text generation and long-context understanding (up to 512K+ tokens) at low memory and cost; intended for consumer electronics, automotive, robotics and enterprise AI developers.

2MiniCPM-V Multimodal Model Series
Scale indicator5 records

Each record includes

Type, Value, Description, Source

Partnership9 partners
Strategic tierCoreTypeStrategic or Co-development PartnerAnnounced on2025-11-01
Description

Strategic partnership with Intel for AI PC optimization. ModelBest's MiniCPM edge models are deeply adapted and accelerated on Intel's latest desktop flagship chips. Partnership provides significant market push and technical acceleration for AI PC applications.

2Baidu Cloud (百度智能云)
Strategic tierCoreTypeStrategic or Co-development PartnerAnnounced on2025-10-30
Description

Strategic cooperation signed October 30, 2025 to jointly build a large model cloud-edge coordination solution, accelerating model deployment in embodied intelligence, intelligent terminals, and edge computing applications. Partnership signed by CEO Li Dahu and COO Lei Shengtao.

modelbest.cn
Strategic tierCoreTypeStrategic or Co-development PartnerAnnounced on2025-10-01
Description

Joint optimization partnership with MediaTek for the next-generation mobile SoC. ModelBest collaborated with MediaTek to tune on-device AI capabilities for the new 3nm Tianji 9400 mobile chip.

Strategic tierCoreTypeStrategic or Co-development PartnerAnnounced on2025-09-27
Description

Strategic cooperation agreement signed September 27, 2025 for deep collaboration on AI large model technology R&D in automotive. Partnership aims to integrate ModelBest's edge large model capabilities with Great Wall Motors' technology stacks across multiple areas including intelligent driving and user services.

Strategic tierCoreTypeStrategic or Co-development PartnerAnnounced on2025-09-19
Description

HUIXI and ModelBest signed strategic cooperation agreement on September 19, 2025 to advance edge AI innovation for automotive and embodied intelligence. The partnership combines HUIXI's high-compute AI chips with ModelBest's MiniCPM large-model series through joint optimization, co-debugging, deployment validation, and co-development. Targets applications in cars, smartphones, PCs, and embodied intelligence systems.

Strategic tierCoreTypeStrategic or Co-development PartnerAnnounced on2025-09-19
Description

Partnership at 2024 World Robot Conference to create a complete 'embodied intelligence' pioneer solution. Industry's first efficient edge model running on humanoid robot demonstration for understanding, reasoning, and physical world interaction.

Strategic tierMinorTypeStrategic or Co-development PartnerAnnounced on2025-09-19
Description

Collaboration to bring super realistic, emotionally nuanced robot pets. Partnership combines cloud and edge models for differentiated robotic product experiences.

8Tsinghua University NLP Lab
Strategic tierCoreTypeStrategic or Co-development PartnerAnnounced on2024-09-25
Description

Ongoing research collaboration with Tsinghua University's NLP laboratory. Joint development of UltraEval-Audio framework, research publications in Nature Communications, and academic沙龙 activities.

modelbest.cn
Strategic tierCoreTypeStrategic or Co-development PartnerAnnounced on2024-07-01
Description

Zhihu partnered with ModelBest to develop zhida.ai, an AI-driven Q&A feature providing users with summarized and detailed responses. ModelBest provided its Zhihaitu AI model for the new Q&A feature deployment.

Recent move7 records

Each record includes

Date, Type, Title, Description, Source

Expansion highlight6 records

Each record includes

Type, Description

Peers3 records
TypeDirect peer
TypeDirect peer
Description

Chinese foundation model developer behind the GLM series. Z.ai co-invested in ModelBest's seed and Series A rounds alongside Zhihu, indicating close competitive and ecosystem overlap in the Chinese LLM market. Both compete for Chinese enterprise, automotive, and government AI deployments.

3Alibaba Qwen Team
TypeBroad incumbent
Description

Developer of the Qwen family of LLMs and multimodal models including on-device variants (Qwen2.5-VL). As a hyperscaler-backed effort with massive compute budget, Qwen is a direct competitor in Chinese edge and cloud AI, and was specifically benchmarked against by ModelBest's MiniCPM-V 4.5.

Market position
Strengths5 records

Each record includes

Headline, Details, Source

Weaknesses5 records

Each record includes

Headline, Details, Source

Competitive moat6 records

Each record includes

Type, Details

Key risks6 records

Each record includes

Headline, Details, Source

Key highlights7 records

Each record includes

Headline, Details, Source

Customer concentration

Classification, Details

Named customers7 records

Each record includes

Name, Industry, Type, Use case, Source, UUID

Segment4 records

Each record includes

Title, Type, Primary, Description, Pain point addressed, Use case, Source

Ideal customer profile5 records

Each record includes

Profile, Firmographic size, Sales motion, Sales cycle length, Buying structure, Purchase trigger, Buyer persona, Geography, Industry vertical, Primary use case, Description, Pain points, Evidence proof points, Target buyer

Technology focused
Yes
API detail
Has APIbool
No

Docs URL, Description

Integration3 records

Each record includes

Title, Type, Description, Source

AI capability12 records

Each record includes

Type, Description, Source

AI maturity
App detail

Has app

Feature11 records

Each record includes

Title, Differentiator, Description, Source

Core technology
Revenue estimate
Valuation estimate
Number of profiles
Profiles4 records

Each record includes

Name, Designation, Designation category, Overview, Profile commentary, Source

No data
Compliance1 record

Each record includes

Name, Class, Description

Funding overview

Funding stage, Last funding date, Total funding USD

Funding rounds11 records

Each record includes

Round, Amount USD, Date, Pre money valuation, Total investors, Investors, News

Investors28 records

Each record includes

Name, Type, Date of entry, Rounds participated, Website

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Each record includes

Name, Acquisition type, Announced date, Completed date, Status, Website, News

Investment1 record

Each record includes

Name, Round, Announced date, Lead investor, Website, News

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

ModelBest

Edge AI Foundation Modelsmodelbest.cn

ModelBest (面壁智能) is a Beijing-based AI company, founded in 2022, that develops lightweight on-device large language, multimodal, and voice models (MiniCPM, MiniCPM-V, MiniCPM-o, VoxCPM) and licenses them to automotive OEMs, chip makers, consumer-electronics manufacturers, robotics firms, and government institutions including China's Supreme People's Court.

What ModelBest does

ModelBest (北京面壁智能科技有限责任公司) is a Beijing-based AI foundation-model company founded in 2022 that develops lightweight, high-performance large language and multimodal models optimized for on-device edge deployment. Its core product family is the MiniCPM series, which spans text LLMs (0.5B to 8B parameters), multimodal vision-language models (MiniCPM-V 2.0 through 4.5), the full-modal MiniCPM-o 2.6, and the VoxCPM voice generation model, all built on a proprietary theoretical framework the company calls the 'Knowledge Density Law' (面壁定律). The technology stack emphasizes sparse attention, int4 quantization, the CPM.cu inference framework, and chip-level co-optimization, enabling models to run on consumer-grade hardware with as little as 2GB of memory at up to 220x inference acceleration. The company open-sources its models through the OpenBMB community (100,000+ GitHub stars, 1M+ downloads) while commercializing through enterprise B2B licensing, OEM/chip integration, and bespoke solutions.

ModelBest monetizes primarily through model licensing and OEM integration fees (MiniCPM embedded in partners' products and platforms such as MediaTek mobile SoCs, Intel AI PC chips, and Baidu Cloud) and through professional services for customized industry solutions, including intelligent cabins, on-device AI assistants, and vertical-domain models (e.g., China's first judicial foundation model co-developed with the Supreme People's Court). Its go-to-market blends a product-led/open-source top-of-funnel with enterprise direct sales and chip-OEM channel partnerships targeting automotive OEMs (Geely, Changan, Great Wall Motors, Volkswagen, Huawei), consumer-electronics manufacturers, cloud platforms, robotics companies (Booster Robotics, Elephant Robot), and government/institutional buyers. The company has raised eight funding rounds between 2022 and April 2026, completed multiple Tier-1 automotive and chip OEM partnerships through 2025, and became the first LLM company in China to pass ASPICE L2 automotive-grade certification.

ModelBest firmographics

Firmographics
Name
ModelBest
Legal name
北京面壁智能科技有限责任公司
Website
https://modelbest.cn
Company type
Private
Founded year
2022
Operating status
Operating
Headcount range
51–100 employees
Short description
ModelBest (面壁智能) is a Beijing-based AI company, founded in 2022, that develops lightweight on-device large language, multimodal, and voice models (MiniCPM, MiniCPM-V, MiniCPM-o, VoxCPM) and licenses them to automotive OEMs, chip makers, consumer-electronics manufacturers, robotics firms, and government institutions including China's Supreme People's Court.
Ownership category
akta.pro rank

ModelBest industry classification

Industry
Product category
Edge AI Foundation Models
NAICS
Software Publishers (5132), Custom Computer Programming Services (541511), Computer Systems Design and Related Services (5415)
SIC
Services-Prepackaged Software (7372), Services-Computer Programming Services (7371), Services-Computer Integrated Systems Design (7373)
akta.pro industry
PRIMARYHDAAACAN
On-Device/Edge Foundation Models (Mobile/Embedded LLMs)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Foundation Models & Generative AI (LLMs, multimodal)
SECONDARYHDAEANAC
Model Serving, Inference & Deployment Platforms (APIs, Edge/On-Prem)
Hardware, Technology & Telecom/Enterprise Platforms & Databases/Artificial Intelligence Platforms for Enterprise
SECONDARYHDAAACAB
Model Hosting, Serving & Inference Platforms
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Foundation Models & Generative AI (LLMs, multimodal)
SECONDARYHDAAACAC
Fine-Tuning, Adaptation & Custom Model Training (PEFT/LoRA/RLHF)
Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Foundation Models & Generative AI (LLMs, multimodal)
SECONDARYBPAEAHAH
MLOps/LLMOps & Model Lifecycle Management Services

Keywords

  • Edge AI models
  • Lightweight large language models
  • On-device AI inference
  • Multimodal foundation models
  • AI model licensing

Where ModelBest is headquartered

Location

Headquarters

HQ city
Beijing
HQ country
China
HQ region
Asia

Offices1 record

Markets served

ModelBest business model

Business model
GTM type
B2B
Offering type
Software
Cost components
Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations

Revenue model

  1. Model Licensing and OEM Integration: License MiniCPM model series to enterprise customers including automotive OEMs, smartphone/PC manufacturers, and robotics companies. Models are integrated into partners' chips (MediaTek SoC), platforms (Intel AI PC, Baidu Cloud), and products. Revenue derived from technology licensing fees and per-unit royalties where applicable.
  2. Enterprise B2B Solutions: Provide customized AI solutions for specific industries such as judicial/legal systems (first judicial judgment vertical domain model in China), automotive (intelligent cabins), and consumer electronics. Revenue from solution development fees and ongoing support contracts.

Go-to-market motion1 record

Distribution channels5 records

Marketing channels6 records

ModelBest product offering

Product offering

Core offering

ModelBest develops and licenses the MiniCPM series of lightweight, high-performance large language and multimodal AI foundation models optimized for on-device (edge) deployment. The portfolio includes text LLMs (MiniCPM 4.0/3.0/2.0/1.0), vision-language models (MiniCPM-V 4.5/2.6/2.5/2.0), full-modal models (MiniCPM-o 2.6), and voice generation (VoxCPM). Models are delivered via open-source community distribution, enterprise licensing to OEMs and chip makers, and integration into consumer electronics, automotive intelligent cabins, embodied robots and government systems.

Product overview

ModelBest is an edge AI company offering a comprehensive portfolio of efficient on-device large language models. The product suite centers on the MiniCPM series for text generation, MiniCPM-V for multimodal (image/video) understanding, MiniCPM-o for full-modal (audio/video/text/image) capabilities, and VoxCPM for voice generation. These models are deployed across consumer electronics (phones, PCs) via Pinea brand, automotive intelligent cabins, embodied robots, and wearable devices. The company pioneered the 'Knowledge Density Law' guiding efficient model development where parameter counts halve every 3.3 months for equivalent performance. Models support major inference frameworks (llama.cpp, ollama, vLLM) and are available through the OpenBMB open-source community.

Differentiator

Problem solved

Functional benefit

Brands

  • MiniCPM (小钢炮): World-leading lightweight and high-performance LLM series for edge devices including phones, PCs, AI cabins, and embodied robots.
  • OpenBMB
  • Pinea (松果派)
  • Lantay (兰台令)
  • VoxCPM

Products and services

  • MiniCPM Series Lightweight, high-performance on-device large language model family in 8B, 4B, 2.4B, 1.2B and 0.5B variants that delivers ChatGPT- to GPT-4o-level text generation and long-context understanding (up to 512K+ tokens) at low memory and cost; intended for consumer electronics, automotive, robotics and enterprise AI developers.
  • MiniCPM-V Multimodal Model Series

Quantifiable outcome

  • MiniCPM 4.0 achieves performance on par with Qwen-3-8B and surpassing Gemma-3-12B with only 22% training cost
  • +6 more outcomes

Companies that use ModelBest

Customer profile

Named customers7 records

Segments4 records

Ideal customer profiles5 records

ModelBest technology and API

Technology

Technology focussed Yes

API detail

Has API
No
API docs
API detail

Core technology

AI maturity

App detail

Integration3 records

AI capability12 records

Feature11 records

ModelBest partnerships and signals

Strategic signal

Partnerships

Nine partnerships are on record, tiered core and minor.

  • IntelcoreStrategic or Co-development Partner · 1 November 2025Strategic partnership with Intel for AI PC optimization. ModelBest's MiniCPM edge models are deeply adapted and accelerated on Intel's latest desktop flagship chips. Partnership provides significant market push and technical acceleration for AI PC applications.
  • Baidu Cloud (百度智能云)coreStrategic or Co-development Partner · 30 October 2025Strategic cooperation signed October 30, 2025 to jointly build a large model cloud-edge coordination solution, accelerating model deployment in embodied intelligence, intelligent terminals, and edge computing applications. Partnership signed by CEO Li Dahu and COO Lei Shengtao.
  • MediaTekcoreStrategic or Co-development Partner · 1 October 2025Joint optimization partnership with MediaTek for the next-generation mobile SoC. ModelBest collaborated with MediaTek to tune on-device AI capabilities for the new 3nm Tianji 9400 mobile chip.
  • Great Wall Motors (长城汽车)coreStrategic or Co-development Partner · 27 September 2025Strategic cooperation agreement signed September 27, 2025 for deep collaboration on AI large model technology R&D in automotive. Partnership aims to integrate ModelBest's edge large model capabilities with Great Wall Motors' technology stacks across multiple areas including intelligent driving and user services.
  • Huixi Technology (HUIXI)coreStrategic or Co-development Partner · 19 September 2025HUIXI and ModelBest signed strategic cooperation agreement on September 19, 2025 to advance edge AI innovation for automotive and embodied intelligence. The partnership combines HUIXI's high-compute AI chips with ModelBest's MiniCPM large-model series through joint optimization, co-debugging, deployment validation, and co-development. Targets applications in cars, smartphones, PCs, and embodied intelligence systems.
  • Booster Robotics (加速进化机器人)coreStrategic or Co-development Partner · 19 September 2025Partnership at 2024 World Robot Conference to create a complete 'embodied intelligence' pioneer solution. Industry's first efficient edge model running on humanoid robot demonstration for understanding, reasoning, and physical world interaction.
  • Elephant Robot (大象机器人)minorStrategic or Co-development Partner · 19 September 2025Collaboration to bring super realistic, emotionally nuanced robot pets. Partnership combines cloud and edge models for differentiated robotic product experiences.
  • Tsinghua University NLP LabcoreStrategic or Co-development Partner · 25 September 2024Ongoing research collaboration with Tsinghua University's NLP laboratory. Joint development of UltraEval-Audio framework, research publications in Nature Communications, and academic沙龙 activities.
  • Zhihu (知乎)coreStrategic or Co-development Partner · 1 July 2024Zhihu partnered with ModelBest to develop zhida.ai, an AI-driven Q&A feature providing users with summarized and detailed responses. ModelBest provided its Zhihaitu AI model for the new Q&A feature deployment.

Scale indicators5 records

Recent moves7 records

Expansion highlights6 records

ModelBest competitors and assessment

Company assessment

Direct peers

  • DeepSeek:
  • Zhipu AI (Z.ai): Chinese foundation model developer behind the GLM series. Z.ai co-invested in ModelBest's seed and Series A rounds alongside Zhihu, indicating close competitive and ecosystem overlap in the Chinese LLM market. Both compete for Chinese enterprise, automotive, and government AI deployments.

Broad incumbents

  • Alibaba Qwen Team: Developer of the Qwen family of LLMs and multimodal models including on-device variants (Qwen2.5-VL). As a hyperscaler-backed effort with massive compute budget, Qwen is a direct competitor in Chinese edge and cloud AI, and was specifically benchmarked against by ModelBest's MiniCPM-V 4.5.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks6 records

Key highlights7 records

Customer concentration

ModelBest social profiles

Digital presence

ModelBest compliance and trust

Trust signal

Compliance1 record

ModelBest financial estimates

Financial estimate

Revenue estimate

Valuation estimate

ModelBest leadership team

Management profile

Number of profiles

Profiles4 records

ModelBest funding detail

Funding detail

Funding overview

Funding rounds11 records

Investors28 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

ModelBest M&A and investment

M&A and investment

M&A

Investments1 record

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about ModelBest

What does ModelBest do?

ModelBest develops and licenses the MiniCPM series of lightweight, high-performance large language and multimodal AI foundation models optimized for on-device (edge) deployment. The portfolio includes text LLMs (MiniCPM 4.0/3.0/2.0/1.0), vision-language models (MiniCPM-V 4.5/2.6/2.5/2.0), full-modal models (MiniCPM-o 2.6), and voice generation (VoxCPM). Models are delivered via open-source community distribution, enterprise licensing to OEMs and chip makers, and integration into consumer electronics, automotive intelligent cabins, embodied robots and government systems.

Is ModelBest a public or private company?

ModelBest is a private company. It is classified as venture growth investor backed and is currently operating.

When was ModelBest founded?

ModelBest was founded in 2022. It employs 51 to 100 people.

Where is ModelBest based?

ModelBest is headquartered in Beijing, China, in the Asia region.

How does ModelBest make money?

Two revenue lines are on record. Model Licensing and OEM Integration is the primary driver. The others are enterprise B2B Solutions.

Who are ModelBest's main competitors?

Direct peers on record are DeepSeek and Zhipu AI (Z.ai). Alibaba Qwen Team is listed as a broad incumbent.

Does ModelBest have an API?

No public API is recorded for ModelBest.

What industry is ModelBest in?

ModelBest's product category is Edge AI Foundation Models. Its primary akta.pro industry code is HDAAACAN, On-Device/Edge Foundation Models (Mobile/Embedded LLMs), with a secondary code of HDAEANAC, Model Serving, Inference & Deployment Platforms (APIs, Edge/On-Prem). Its NAICS code is 5132 and its SIC code is 7372.

Unlock the full company data

50 free credits on sign-up, no credit card required.

Contact sales
Live signals
PandailyModelBest Details LES, the Method Behind a 35B MoE Model Running on Phones at Over 20 Tokens/sModelBest published an engineering write-up on Lookahead Expert Scheduling (LES), an inference system for its MiniCPM-V 4.7 35B-A3B MoE model. LES prefetches next-layer experts from phone flash, achieving 20.51 tokens per second on an iPhone 18 Pro Max, a 6.07x speedup over baseline. The company has open-sourced the code.Laotian TimesModelBest and OpenBMB Open-Source MiniCPM-OCR Series, Delivering High-Performance Document Parsing and AI-Powered Quality ControlModelBest and OpenBMB released the MiniCPM-OCR series, including MiniCPM-OCR 1.0 (1.3B) and Jev (2B), for document parsing and quality control. The models support edge deployment, bilingual content, and key information extraction, with Jev acting as an OCR result inspector. They are open-sourced on GitHub, Hugging Face, and ModelScope.PR Newswire APACModelBest and OpenBMB Open-Source MiniCPM-OCR Series, Delivering High-Performance Document Parsing and AI-Powered Quality ControlModelBest and OpenBMB released the MiniCPM-OCR series, including MiniCPM-OCR 1.0 (1.3B) and Jev (2B), for document parsing and quality control. The models support edge deployment, bilingual content, and key information extraction, with Jev acting as an AI quality inspector. They are available on GitHub, Hugging Face, and ModelScope, with Day-0 hardware support.PR NewswireModelBest and OpenBMB Open-Source MiniCPM-OCR Series, Delivering High-Performance Document Parsing and AI-Powered Quality ControlModelBest and OpenBMB released the MiniCPM-OCR series, including MiniCPM-OCR 1.0 (1.3B) and Jev (2B), for document parsing and quality control. The models support edge deployment, bilingual content, and key information extraction, with Jev acting as an OCR result inspector. They are open-sourced on GitHub, Hugging Face, and ModelScope.AdvfnModelBest and OpenBMB Open-Source MiniCPM-OCR Series, Delivering High-Performance Document Parsing and AI-Powered Quality ControlModelBest and OpenBMB released the MiniCPM-OCR series, including MiniCPM-OCR 1.0 (1.3B) and Jev (2B), for document parsing and quality assessment. The models support edge deployment, bilingual content, and key information extraction, with Jev acting as an AI quality inspector. They are available on GitHub, Hugging Face, and ModelScope.PandailyTsinghua, ModelBest and OpenBMB Release PilotDeck 2.0, an Open-Source Agent Operating SystemTsinghua, ModelBest, OpenBMB, and AI9Stars released PilotDeck 2.0, an open-source agent operating system with isolated workspaces and editable memory. Team tests cut long-task costs by about 83%, and WorkBuddy Bench scores improved to 79.1 with Qwen3.6-35B-A3B.PR Newswire APACModelBest and OpenBMB Launch MiniCPM-V 4.7, Bringing a 35B Multimodal MoE Model to Phones and Tablets With About 3 GB of MemoryModelBest and OpenBMB released MiniCPM-V 4.7, a 35B multimodal MoE model for on-device use with about 3GB memory. It runs on iOS and Android, achieving over 17 tokens per second on an iPad Pro M4. The model supports 4× and 16× visual token compression and has Day-0 support from major chipmakers.PR Newswire APACModelBest and OpenBMB Launch MiniCPM-V 4.7, Bringing a 35B Multimodal MoE Model to Phones and Tablets With About 3 GB of MemoryModelBest and OpenBMB released MiniCPM-V 4.7, a 35B multimodal MoE model for on-device use with about 3GB memory. It runs on iOS and Android, achieving over 17 tokens per second on an iPad Pro M4. The model supports 4× and 16× visual token compression and has Day-0 support from major chipmakers.PR NewswireModelBest and OpenBMB Launch MiniCPM-V 4.7, Bringing a 35B Multimodal MoE Model to Phones and Tablets With About 3 GB of MemoryModelBest and OpenBMB released MiniCPM-V 4.7, an open-source on-device multimodal model in 35B-A3B and 1B versions. The 35B-A3B version activates about 3 billion parameters per token and runs on devices with about 3GB memory, achieving over 17 tokens per second on an iPad Pro M4. It supports 4× and 16× visual token compression with strong benchmark scores.PandailyModelBest Open-Sources MiniCPM-V 4.7, Its First MoE Multimodal Model, Running in 3GB on a PhoneModelBest released MiniCPM-V 4.7, an open-source multimodal model in 35B-A3B and 1B sizes, with the 35B-A3B being the first MoE variant. It runs on an iPhone 17 Pro Max using about 3GB of memory, decoding over 13 tokens per second. The model scores 91.72 on MMBench and 90.21 on CMMLU.