Instructions to use spicyneuron/Kimi-K2.6-MLX-3.6bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use spicyneuron/Kimi-K2.6-MLX-3.6bit with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("spicyneuron/Kimi-K2.6-MLX-3.6bit")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps
LM Studio

Pi new

How to use spicyneuron/Kimi-K2.6-MLX-3.6bit with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "spicyneuron/Kimi-K2.6-MLX-3.6bit"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "spicyneuron/Kimi-K2.6-MLX-3.6bit"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use spicyneuron/Kimi-K2.6-MLX-3.6bit with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "spicyneuron/Kimi-K2.6-MLX-3.6bit"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default spicyneuron/Kimi-K2.6-MLX-3.6bit

Run Hermes

hermes

MLX LM

How to use spicyneuron/Kimi-K2.6-MLX-3.6bit with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "spicyneuron/Kimi-K2.6-MLX-3.6bit"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "spicyneuron/Kimi-K2.6-MLX-3.6bit"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "spicyneuron/Kimi-K2.6-MLX-3.6bit",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Will there be a 2.8bit quant?

by darthsider - opened 26 days ago

Discussion

darthsider

26 days ago

spicyneuron/Kimi-K2.5-MLX-2.8bit fits well on M3 Ultra 512 GB with full context. So it’s the perfect size. Would you be doing 2.8bit for Kimi-K2.6 as well?

spicyneuron

Owner 26 days ago

•

edited 26 days ago

Stay tuned! Kimi trials take a bit longer since I can only fit 3 versions on a 4 TB external SSD (the dequantized model alone is 2 TB+).

So far, experiments in the 2.9 - 3.3 bpw range had significantly higher KL divergence. Still searching for the best tradeoff.

darthsider changed discussion status to closed 26 days ago

darthsider

26 days ago

much appreciated

spicyneuron

Owner 25 days ago

Uploading a 430 GB version now: https://huggingface.co/spicyneuron/Kimi-K2.6-MLX-3.3bit

This was a tricky one. At 2.9 bits, K2.6 is still ~400 GB. Perplexity barely moves, but KL divergence triples, and other evals noticeably decay.

It's entirely possible my Kimi K2.5 quants had similar decay, but my earlier workflows didn't capture it. In any case, let me know how it runs!

spicyneuron changed discussion status to open 25 days ago

darthsider

25 days ago

I’m gonna check it out. Thanks.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment