# FlagOpen/RoboBrain2.5

RoboBrain 2.5: Advanced version of RoboBrain. Depth in Sight, Time in Mind. 🎉🎉🎉

Repository: https://github.com/FlagOpen/RoboBrain2.5
Canonical: https://ross.abutalabs.com/products/robobrain25
Homepage: https://superrobobrain.github.io/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: embodied-ai, multimodel-large-language-model
Last push: 2026-02-28T20:04:25+00:00

## Health v2 (maintenance only)
Score: 50/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 69, release rhythm 35, longevity 32
- inputs: {"age_days": 454, "days_push": 186, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1132, forks 115 (observed 2026-08-28T04:03:42.621270+00:00)

## What it is
RoboBrain 2.5 is an open-source embodied AI foundation model from BAAI that combines multimodal large language model capabilities with 3D spatial reasoning and dense temporal value estimation for robotic manipulation. The repository provides model weights, inference code, and training resources for depth-aware perception and execution-aware embodied intelligence.

## Use cases
- run an embodied AI model for robotic manipulation planning
- predict 3D keypoint trajectories for pick-and-place tasks
- estimate depth-aware spatial coordinates from camera images
- model step-wise task progress for robot execution feedback
- fine-tune a multimodal LLM for embodied reasoning
- research spatial grounding and metric measurement in robotics

## When to choose
- you need an open embodied foundation model with 3D spatial reasoning
- you want dense temporal progress estimation for robot learning
- you are doing research on vision-language models for manipulation

## When to avoid
- you need a lightweight model for CPU-only deployment
- you need a production-ready turnkey robot control stack
- your task is unrelated to embodied or spatial AI

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, computer-vision, nlp, llm-inference
- domain: artificial-intelligence, machine-learning, robotics, computer-vision, large-language-models
- platform: python
- tags: embodied-ai, multimodal-llm, foundation-model, spatial-reasoning, robotics, temporal-modeling, model-weights, gpu, linux

## Member repositories
- FlagOpen/RoboBrain2.5 (main) score 50

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:42.621270+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:37:55.281400+00:00, confidence not recorded.
  - readme: https://github.com/FlagOpen/RoboBrain2.5 (fetched 2026-08-28T04:03:42.621270+00:00, sha f83a918e2971)
  - homepage: https://superrobobrain.github.io/ (fetched 2026-08-29T12:42:26.807768+00:00, sha 32d62856b4f6)
- Data as of 2026-08-30T08:39:29.467469+00:00.