# MoonshotAI/Kimi-K2.5

Open Visual Agentic Intelligence

Repository: https://github.com/MoonshotAI/Kimi-K2.5
Canonical: https://ross.abutalabs.com/products/kimi-k25
License: NOASSERTION
License Family: other
Last push: 2026-08-06T10:42:43+00:00

## Health v2 (maintenance only)
Score: 58/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 35, longevity 15
- inputs: {"age_days": 215, "days_push": 27, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2303, forks 324 (observed 2026-08-28T04:06:35.611563+00:00)

## What it is
Kimi K2.5 is Moonshot AI's open-weight, 1T-parameter Mixture-of-Experts multimodal agentic model (32B activated), continually pretrained on ~15T mixed visual and text tokens. The repository provides model weights, documentation, and resources for its vision-language understanding, code generation from visual specs, and multi-agent 'Agent Swarm' execution capabilities.

## Use cases
- run an open multimodal agentic llm locally
- generate code from ui designs or screenshots
- build multi-agent workflows that decompose tasks into sub-agents
- visual question answering and cross-modal reasoning
- agentic tool use grounded in visual inputs
- download kimi k2.5 weights from hugging face

## When to choose
- you need an open-weight frontier model with native vision-language and agentic capabilities
- you want a single model that handles both instant and thinking modes
- you need parallel multi-agent task decomposition out of the box

## When to avoid
- you lack the GPU infrastructure to serve a 1T-parameter MoE model
- you need a small, lightweight model for edge or on-device use
- you need a permissively licensed model for commercial redistribution without checking the modified MIT terms

## Facets
- artifact type: learning-resource
- maturity: active
- function: llm-inference, agent-framework, machine-learning, nlp, computer-vision
- domain: large-language-models, artificial-intelligence, computer-vision
- platform: python
- tags: mixture-of-experts, multimodal-llm, open-weights, agentic-ai, vision-language-model, moonshot-ai, kimi, model-weights, ai-agents, multimodal, gpu, linux

## Member repositories
- MoonshotAI/Kimi-K2.5 (main) score 58

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:35.611563+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 2026-08-30T02:39:53.786115+00:00, confidence not recorded.
  - readme: https://github.com/MoonshotAI/Kimi-K2.5 (fetched 2026-08-28T04:06:35.611563+00:00, sha d8aa5ba0cf53)
- Data as of 2026-08-30T08:39:29.467469+00:00.