Parameters
—
Weights (est.)
—
Mac memory (est.)
—
Size is estimated from the parameter count and quantisation. Memory allows a fifth extra for context plus 2 GB for the system; long contexts need more.
Run it
With mlx-vlm on an Apple silicon Mac:
pip install mlx-vlmpython -m mlx_vlm.generate --model mlx-community/Kimi-K2.6-MoE-Smart-Quant --image photo.jpg --prompt "Describe this image"
Details
- Base model
- moonshotai/Kimi-K2.6 (quantized) · Kimi K2.6 on the radar →
- Vendor
- Moonshot · Kimi
- Task
- Vision + text
- Quantisation
- Not stated
- Licence
- —
- Downloads
- 0 in the last 30 days · 0 all time · 16 likes
- Published
- 2026-04-20 · updated 2026-04-20 (6mo ago)
- Languages
- en
- Tags
- kimi_k25, quantized, moe-aware-quant, custom_code
Other MLX builds of moonshotai/Kimi-K2.6
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Kimi-K2.6-mlx-DQ3_K_M-q8 | 8-bit | ≈ 1091 GB | > 512 GB | 933 |
| Kimi-K2.6-MoE-Smart-Quant (this one) | — | — | — | 0 |