ByteDance-Seed/m3-agent
A multimodal agent framework from ByteDance that processes visual and auditory inputs, builds episodic and semantic long-term memory, and performs autonomous multi-turn reasoning.

M3-Agent is a multimodal agent system that processes real-time visual and auditory inputs to build and update long-term memory. It combines episodic memory for experience-based learning with semantic memory for world knowledge accumulation. The agent organizes memory in an entity-centric multimodal format and performs iterative reasoning with retrieved information. The repository includes M3-Bench, a benchmark with 1000 real-world and web videos for evaluating memory effectiveness and multimodal reasoning in agents.
Frequently asked
- What is ByteDance-Seed/m3-agent?
- A multimodal agent framework from ByteDance that processes visual and auditory inputs, builds episodic and semantic long-term memory, and performs autonomous multi-turn reasoning.
- Is m3-agent open source?
- Yes — ByteDance-Seed/m3-agent is open source, released under the Apache-2.0 license.
- What language is m3-agent written in?
- ByteDance-Seed/m3-agent is primarily written in Python.
- How popular is m3-agent?
- ByteDance-Seed/m3-agent has 1.4k stars on GitHub.
- Where can I find m3-agent?
- ByteDance-Seed/m3-agent is on GitHub at https://github.com/ByteDance-Seed/m3-agent.