Catalog Systems Dedicated memory layers M3-Agent (ByteDance)

M3-Agent (ByteDance)

https://github.com/ByteDance-Seed/m3-agent

ByteDance Seed + Zhejiang Univ + SJTU. Multimodal agent with long-term memory as entity-centric graph built from live video + audio streams. Parallel memorisation + control processes. Ships M3-Bench evaluation dataset.

Active · Verified 2026-10-22 · fresh as of 2026-10-22 · 37 cited facts, 2 with a third-party source · Suggest a correction

At a glance

Type
Episodic + semantic multimodal memory graph repo fix
Tier
T2
Section
Dedicated memory layers
Created
2025-07 repo fix
Latest release
no releases fix
License
Apache-2.0 repo fix
GitHub
ByteDance-Seed/m3-agent — ~1.4k stars, last commit not confirmed repo fix
Pricing
not applicable — academic research code (no commercial pricing) repo fix
Funding
not applicable — ByteDance corporate research repo fix
Freshness
2026-10-22 fix
Layer
agent-infra repo fix
Licensing
open-source repo fix
Deployment locus
local repo fix
Platform vs product
platform repo fix
SDK languages
python repo fix
Developer activity
OpenRank 1.05 (12-mo avg 1.58) · 1 participants — 2025-12 third-party fix
Maintainer health
0 issues in 90 days · 0 releases in 12 months · bus factor 1 — as of 2026-10-31 · concern: one contributor wrote 100% of sampled commits repo

Memory taxonomy

storage
graph
retrieval
graph-traversal
persistence
long-term
update
extraction
unit
episode
governance
opaque
conflict
n/a

When to use

Optimised for: episodic + semantic multimodal memory graph (research)

Anti-fit: not applicable - research paper

Pros & cons

Pros

Multi-modal multi-memory agent — three coordinated memory streams for video, audio, text agent tasks.

Cons

ByteDance-internal; published architecture but production access limited; eval primarily on Chinese benchmarks.

Claims & capabilities

+6.7% / +7.7% / +5.3% accuracy on M3-Bench-robot / M3-Bench-web / VideoMME-long vs Gemini-1.5-pro + GPT-4o baseline.

Technical surface

API surface
searched not found
Backend storage
searched not found
Deployment
Self-hosted (research code; API config + memory graph files)
Embedding model
searched not found
Multi-tenancy
searched not found
MCP
not documented publicly
A2A
not documented publicly
OpenTelemetry
not documented publicly

Alternatives to M3-Agent (ByteDance) → live options ranked by shared design, licence, hosting and MCP support

Similar systems

Other dedicated memory layers in the catalog, ranked by inbound references.

  • Mem0 T1

    Universal memory layer for AI agents. Three concurrent stores (vector + graph + KV); LLM-extracted facts; concurrent retrieval via ThreadPoolExecutor.

  • Zep & Graphiti T1

    Bi-temporal knowledge graph (event time + ingestion time). Strong on chronological reasoning and contradiction tracking. Graphiti is the open-source core.

  • Cognee T1

    "Extract–Cognify–Load" pipeline that turns raw input into a typed, queryable knowledge graph for agent recall.

  • Hindsight (Vectorize) T1

    Standalone memory service from Vectorize. Open source. Biomimetic four-network design (World, Bank, Observation, Opinion). Ships an MCP memory server.

  • Letta / MemGPT T1

    Originated as MemGPT (UC Berkeley), where the model pages its own memory between context, recall and archival tiers. The product now centres on Letta Code, a stateful agent harness whose agents rewrite their own memory, skills and prompts over time.

  • Memori T2

    Structured, persistent memory layer for AI agents. Open source.

Sources (36 cited facts, 2 third-party)

Each fact links to the page it came from; the badge says who published it (how to read sources).