OpenBMB

Text-to-speech spanning 30 languages, with voice cloning and new voices designed from text descriptions.

Pocket-sized multimodal LLM for efficient image and video understanding on phones and edge devices.

Low-code MCP framework from OpenBMB for assembling RAG pipelines from retrieval, rerank, and generation parts.

Framework from OpenBMB for deploying multiple LLM-based agents across simulation, task solving, and collaboration.

Family of efficient end-side LLMs from OpenBMB that run on phones and edge devices with strong reasoning per parameter.

OpenBMB platform for training, serving, and evaluating tool-learning LLMs on a set of over 16K real-world APIs.

OpenBMB vision-language-action model for robotics that pairs reinforcement learning with explicit reasoning.

End-to-end infrastructure for training and evaluating LLM agents, covering the lifecycle from data to eval.

Framework in which distributed AI agents discover one another and collaborate over an internet-like protocol.

Router that fans Claude-style requests across multiple LLM backends, a drop-in for Anthropic-API clients.

RAG pipeline built on vision-language models that indexes PDFs and slides as images, skipping text parsing.

General-purpose mobile agent that uses multimodal foundation models for long-horizon tasks in phone apps.

Reference apps showing how to run MiniCPM-V multimodal models on-device, as a starting point for vision apps.

Open-source framework for evaluating speech models across recognition, speech-to-text, and speech-to-speech tasks.

LLM-powered agent that explores a repository and auto-generates documentation for faster team onboarding.

Multi-level memory plugin that gives OpenClaw agents long-term context across sessions.

Benchmark toolkit that measures how well retrieval-augmented generation pipelines answer questions.

Open source framework that evaluates foundation models through modular benchmarks for capability assessment.