Instructions to use jedisct1/Ornith-1.0-35B-MLX-oQ8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jedisct1/Ornith-1.0-35B-MLX-oQ8 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("jedisct1/Ornith-1.0-35B-MLX-oQ8")

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 jedisct1/Ornith-1.0-35B-MLX-oQ8 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "jedisct1/Ornith-1.0-35B-MLX-oQ8"

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": "jedisct1/Ornith-1.0-35B-MLX-oQ8"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use jedisct1/Ornith-1.0-35B-MLX-oQ8 with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "jedisct1/Ornith-1.0-35B-MLX-oQ8"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "jedisct1/Ornith-1.0-35B-MLX-oQ8"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "jedisct1/Ornith-1.0-35B-MLX-oQ8",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use jedisct1/Ornith-1.0-35B-MLX-oQ8 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 "jedisct1/Ornith-1.0-35B-MLX-oQ8"

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 jedisct1/Ornith-1.0-35B-MLX-oQ8

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use jedisct1/Ornith-1.0-35B-MLX-oQ8 with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "jedisct1/Ornith-1.0-35B-MLX-oQ8"

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 "jedisct1/Ornith-1.0-35B-MLX-oQ8" \
  --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"

Ornith-1.0-35B MLX oQ8

This is a full-width MLX oQ8 build of deepreinforce-ai/Ornith-1.0-35B at revision 5df2ed3f675c7beaa490328cc70bb573b65fb660.

It is not the DOTResize variant. The MoE intermediate width remains 512, with 40 text layers and 256 experts. The artifact is quantized with oMLX affine oQ8 group-size-64 quantization and keeps the 262,144-token context metadata.

The tokenizer includes the chat template directly in tokenizer_config.json, and tool_parser_type is set to qwen3_coder for OpenAI-compatible tool-call parsing in runtimes that use that metadata. Thinking is disabled by default in the template unless a runtime explicitly supplies enable_thinking=true.

Compatibility

This package was checked with both oMLX and LM Studio.

oMLX: loaded as Ornith-1.0-35B-MLX-oQ8, advertised max_model_len = 262144, enabled the qwen3_coder tool parser, and passed a direct OpenAI-compatible tool-call smoke test.
LM Studio: indexed as ornith-1.0-35b-mlx-oq8 with 262,144 max context, trainedForToolUse = true, and 8-bit quantization. It loaded at 262,144 context and answered a local OpenAI-compatible API smoke request.

The config uses vision_config.model_type = qwen3_5_moe, matching the current LM Studio MLX backend support while preserving the Qwen3.5-MoE/VLM metadata shape used by oMLX.

Tool-Calling Validation

The final strict Swival tool-calling suite passed 5/5 at 262,144 context with zero failed tool calls. The held-out direct OpenAI-compatible tool smoke passed 3/3.

The strict Swival system prompt used for the 5/5 run is included as swival_strict_system_prompt.txt so the result can be reproduced without hidden local prompt state.

The accepted strict suite covered:

line-specific edit after reading the file
grep, targeted edit, and test command execution
batch file reads via read_multiple_files
read-before-delete behavior
argv-form command execution

Artifact Notes

Format: MLX safetensors
Quantization: oQ8, affine, group size 64
Context metadata: 262,144 tokens
MTP: not included
Source model: deepreinforce-ai/Ornith-1.0-35B

This is a derived MLX quantized artifact, not a new base model.

Downloads last month: 55

Safetensors

Model size

10B params

Tensor type

BF16

U32

MLX

Hardware compatibility

8-bit

Model tree for jedisct1/Ornith-1.0-35B-MLX-oQ8

Base model

ornith-ai/Ornith-1.0-35B

Quantized

(185)

this model

Collection including jedisct1/Ornith-1.0-35B-MLX-oQ8

Ornith

Collection

3 items • Updated Jun 27