Instructions to use ukisai/Swift-1.5-3bit-MLX-TextOnly with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ukisai/Swift-1.5-3bit-MLX-TextOnly 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("ukisai/Swift-1.5-3bit-MLX-TextOnly")

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 Settings
LM Studio

How to use ukisai/Swift-1.5-3bit-MLX-TextOnly with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "ukisai/Swift-1.5-3bit-MLX-TextOnly"

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "ukisai/Swift-1.5-3bit-MLX-TextOnly"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use ukisai/Swift-1.5-3bit-MLX-TextOnly with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "ukisai/Swift-1.5-3bit-MLX-TextOnly"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "ukisai/Swift-1.5-3bit-MLX-TextOnly"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "ukisai/Swift-1.5-3bit-MLX-TextOnly",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use ukisai/Swift-1.5-3bit-MLX-TextOnly 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 "ukisai/Swift-1.5-3bit-MLX-TextOnly"

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 ukisai/Swift-1.5-3bit-MLX-TextOnly

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ukisai/Swift-1.5-3bit-MLX-TextOnly with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "ukisai/Swift-1.5-3bit-MLX-TextOnly"

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "ukisai/Swift-1.5-3bit-MLX-TextOnly" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Website • Learn more • BF16 model • GGUF • GSQ-RCO GGUF • Evaluation • Enterprise licensing

Swift 1.5 Qwen3.8-27B — 3-bit MLX TextOnly

MLX affine 3-bit quantization, group size 64. Swift 1.5 is UkisAI's reasoning-efficient Qwen3.8-27B derivative, focused on long-horizon, agentic and coding tasks. This export retains the text model; vision and MTP are intentionally omitted. It is not an image/video model or a speculative-MTP implementation.

Swift 1.5 uses 58.5% fewer thinking tokens than base Qwen3.8-27B while scoring 0.35% higher, for a 9.18× speed-up on several tasks.

Use only a complete snapshot whose files match UPLOAD_MANIFEST.json. A build-integrity check is not a generation or quality certificate. BF16 quality parity and long-context operation have not been independently established.

Demo

We gave base Qwen3.8-27B and Swift 1.5 27B the same prompt:

create a 3d little planet globe where I (player can walk around) and it has all these biomes to explore, the globe doesn't have to be too big, but still fun to go around. It's about a boy scout who is camping and goes around exploring.

Try the game yourself here: https://ukisai.com/swift-games/27b

Base Qwen3.8-27B took 104.6 minutes to build its game. Swift 1.5 took 11.39 minutes.

Source and quantization

The recorded conversion source is the customized Swift BF16 model at 33b0d0c52004c08c88a98bb226e07852b9871aae, using official Apple MLX-LM commit c69d1288440a0dc4e6401fc417098b07598dccd5. The later complete BF16 revision 5ad04445d2686f525e9fbe5c077e6fa0c7df4200 is a reference link, not a replacement for the actual conversion provenance. This independent audit has not established full source-value equivalence between those revisions. No base Qwen or other quantized checkpoint is substituted.

The original build contains three shards and 11,771,132,928 bytes of tensor data: 1,847 saved tensors, including 498 packed UINT32 weights and 1,349 BF16 tensors. The original build report records 498 quantized modules and no vision/MTP tensors. The independent recovery check verified full file SHA-256 values and header/index consistency of the original build. These checks do not measure generation quality.

Evaluation

See the Swift BF16 source evaluation. Those results were not rerun on this 3-bit export and are not its benchmark scores. Aggressive 3-bit quantization may reduce quality; no quality-parity claim is made.

Validation and use

USAGE.md provides pinned installation, integrity verification and a text-generation example. This TextOnly export uses the official text architecture; the separate full-model patch that permits only 4/5-bit is not required or claimed to support 3-bit.

QUANTIZATION_MANIFEST.json is the preserved original build report. Its VALIDATED label refers to build/structural checks: its recorded generation test was skipped and contains no generated cases. Do not interpret it as an Apple Silicon, Linux generation, long-context or quality PASS. Package checks separately record the newly performed tokenizer and original-file checks, including their limited scope.

The approximately 11.77 GB tensor payload is not the complete memory requirement. Runtime, cache and OS memory are additional. Do not raise system memory limits to fit this model, or assume long-context support from the tensor size alone. Only text messages are supported. Images, video and MTP are outside this release.

Exact-output formatting is not guaranteed: JSON may include Markdown fences, and Unicode normalization or letter case may change. Validate structured output in your application.

License and access

Swift 1.5 derives from Qwen3.8-27B (Copyright 2026 Alibaba Cloud, Apache License 2.0). UkisAI's adapted weights are licensed under the Swift Open License v1.0. See NOTICE for attribution and change notices.

Personal, research, educational, evaluation and commercial use are free for individuals and organizations with gross annual revenue, including affiliates, of up to US$1,000,000. Above that threshold, commercial use requires a separate Swift Enterprise License. Contact UkisAI for terms. Nothing in the Swift Open License limits the Apache 2.0 rights in Qwen3.8-27B itself. The upstream Apple MLX-LM MIT notice is separate from the model-weight licenses.

Citation

@misc{swift-1.5-qwen3.8-27b,
  title  = {Swift 1.5 Qwen3.8-27B},
  author = {UkisAI},
  year   = {2026},
  url    = {https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b}
}

Acknowledgements

We acknowledge the NVIDIA Innovation Lab, Amazon Web Services, and Google Cloud for compute credits and infrastructure support for Swift's development, training and evaluation.