available
Weights
Official model weights are downloadable.
Inspect this claim’s source ↗
Moonshot AI / FAMILY RECORD
A trillion-parameter visual agentic model with a comparatively small active footprint and public weights and code.
Custom or artifact-specific terms: read them. Generally permitted, with an additional display requirement for products above the license’s stated revenue or user thresholds.
Modified MIT License ↗1T total / 32B active, plus a 400M vision encoder
About 595.18 GB of selected documented weights · File listing ↗
Selected artifact: moonshotai/Kimi-K2.5. Format: Safetensors, native INT4. File evidence checked 2026-10-11. Selected artifact license: modified-mit ↗
All 64 native INT4 checkpoint shards. Modified MIT, not unqualified MIT. Selected weight files only; supporting files and runtime memory are separate. Not a smallest-download census or proof of hardware compatibility.
Compatibility, runtime memory and speed are untested by Bright.
Choose your computer and inspect documented candidates →Source check: . Downloads, when known, carry their own check date.
Checked release: Kimi-K2.5
Sources checked 2026-09-19
K2.5 remains the reference checkpoint here; Moonshot AI’s cited documentation notes a later K2.6 release.
The profile describes this checkpoint, not every release carrying the family name. Developer documentation is not independent reproduction.
THE MODEL / THE TERMS
Specifications are developer-reported. Primary release documentation ↗ · All sources and terms
NINE DIMENSIONS / NO OPENNESS SCORE
Each component has its own evidence. “Not established” means this source review did not verify availability; it is not a claim that nothing exists.
available
Official model weights are downloadable.
Inspect this claim’s source ↗available
Configuration and architecture details are public.
Inspect this claim’s source ↗available
The official repository provides code and deployment examples.
Inspect this claim’s source ↗not established
Complete training code was not confirmed.
Inspect this claim’s source ↗partial
The paper describes training and agentic methods without a complete reproducible recipe.
Inspect this claim’s source ↗partial
Training strategy is described without release of the full mixture.
Inspect this claim’s source ↗not established
No sufficiently specific public artifact was confirmed in this review.
Inspect this claim’s source ↗partial
Vendor evaluation protocols and results are published.
Inspect this claim’s source ↗restricted
The modified MIT license adds a display requirement at defined scale thresholds.
Inspect this claim’s source ↗WHY THIS RELEASE MATTERS
WHAT THE RECORD DOES NOT ESTABLISH
No independent reproduction is attached; benchmark statements remain attributed to Moonshot AI.
FOLLOW THE RECORD
Source check: 2026-09-19. AI-assisted editorial review, with no named external expert or independent hardware validation claimed. Licenses and model cards can change. Review the linked terms for the specific artifact and intended use.
Understand open source vs open weights ↗