Text Generation
MLX
Safetensors
English
Chinese
kimi_k3
axquant
Mixture of Experts
kimi
experimental
not-certified
custom_code
2-bit
Instructions to use AutomatosX/AX-Kimi-K3-MLX-AXQ-2bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use AutomatosX/AX-Kimi-K3-MLX-AXQ-2bit with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # if on a CUDA device, also pip install mlx[cuda] # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("AutomatosX/AX-Kimi-K3-MLX-AXQ-2bit") prompt = "Once upon a time in" text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- MLX LM
How to use AutomatosX/AX-Kimi-K3-MLX-AXQ-2bit with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Generate some text mlx_lm.generate --model "AutomatosX/AX-Kimi-K3-MLX-AXQ-2bit" --prompt "Once upon a time"
- Atomic Chat
AX-Kimi-K3-MLX-AXQ-2bit
Experimental AXQ 2-bit MLX pack of
moonshotai/Kimi-K3.
Not certified. Will not be certified in this revision. Stream required.
Hobby / curiosity only. 2-bit only — no AXQ MXFP4 sibling (source is already
native MXFP4 QAT). Official card has no MTP; this leaf has no -MTP.
Language path only. Vision (MoonViT-V2) stays BF16. Convert requires public mlx-vlm Kimi Delta Attention.
| Item | Status |
|---|---|
Convert + ax_expert_stream.json |
Factory job on df-macstudio-m2 |
| Convert git SHA | 2f4e4e49b82463ac9e146090020a9565c8583253 |
| License | Kimi K3 License copied into the pack |
- Downloads last month
- 436
Model size
349B params
Tensor type
F32
·
U32 ·
Hardware compatibility
Log In to add your hardware
2-bit
Model tree for AutomatosX/AX-Kimi-K3-MLX-AXQ-2bit
Base model
moonshotai/Kimi-K3