moonshotai/Kimi-VL-A3B-Thinking
Parameters16.4B (3B active)
Context128K
Input → outputimage + text → text
Licensemit
PrecisionBF16
ArchitectureKimiVLForConditionalGeneration
ReleasedApr 9, 2025
Downloads14K
Based on moonshotai/Kimi-VL-A3B-Instruct
Other sizes and formats in the family
| Model | Params | Precision | Downloads |
|---|---|---|---|
| moonshotai/Kimi-VL-A3B-Instruct | 16.4B | BF16 | 143.3K |
| moonshotai/Kimi-VL-A3B-Thinking-2506 | 16.4B | BF16 | 7.5K |
This release in the news
No coverage of this exact release yet.
Recent news about Kimi
- Oct 19, 2026 · Tutorial / explainerThe Model Swap Workshop: GPT-5.5 vs Kimi vs DeepSeek vs Mistral
- Oct 17, 2026 · Opinion / analysisAnimLang, Built simple workflow/repo for creating animation using LLM (GSAP, ThreeJS, and Manim)
- Oct 17, 2026 · Tutorial / explainerLocal and Distributed Large LLM Inference with MLX and RDMA on Mac Studio
- Oct 16, 2026 · Opinion / analysisKimi K3, three months after the hype: too big for almost anyone to run locally, and smaller models have already caught it
- Oct 15, 2026 · Opinion / analysis4x R9700 builds with $127 used EPYCs: $10.7k all-in per machine. I built three (384GB VRAM)
- Oct 14, 2026 · Opinion / analysisMapping the Post-Transformer Architecture Landscape in 2026 [D]
- Oct 13, 2026 · Model releaseZ.ai GLM-Legal Beats Claude Fable 5 and Kimi K3 on Harvey LAB
- Oct 12, 2026 · Research paperReplaySSM: Rethinking State Storage for Faster Hybrid-Model Decoding