Instructions to use paxlabs-llm/Matrix.o1-Ornith-1.0-397B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use paxlabs-llm/Matrix.o1-Ornith-1.0-397B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="paxlabs-llm/Matrix.o1-Ornith-1.0-397B")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("paxlabs-llm/Matrix.o1-Ornith-1.0-397B")
model = AutoModelForMultimodalLM.from_pretrained("paxlabs-llm/Matrix.o1-Ornith-1.0-397B", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use paxlabs-llm/Matrix.o1-Ornith-1.0-397B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "paxlabs-llm/Matrix.o1-Ornith-1.0-397B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "paxlabs-llm/Matrix.o1-Ornith-1.0-397B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/paxlabs-llm/Matrix.o1-Ornith-1.0-397B

SGLang

How to use paxlabs-llm/Matrix.o1-Ornith-1.0-397B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "paxlabs-llm/Matrix.o1-Ornith-1.0-397B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "paxlabs-llm/Matrix.o1-Ornith-1.0-397B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "paxlabs-llm/Matrix.o1-Ornith-1.0-397B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "paxlabs-llm/Matrix.o1-Ornith-1.0-397B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use paxlabs-llm/Matrix.o1-Ornith-1.0-397B with Docker Model Runner:
```
docker model run hf.co/paxlabs-llm/Matrix.o1-Ornith-1.0-397B
```

Matrix.o1-Ornith-1.0-397B

Today, PaxLabs is begining work on Matrix.o1-Ornith-1.0-397B, a specialized variant of the Ornith-1.0 family, tuned for seamless integration with the Matrix Agentic System.

Attribution & Heritage

This model is built upon the exceptional foundation of Ornith-1.0-397B by DeepReinforce AI — a state-of-the-art self-improving open-source model family for agentic coding. We extend our deepest gratitude to the DeepReinforce team for their groundbreaking work and for releasing Ornith under the permissive MIT license.

"Standing on the shoulders of giants." — The Ornith architecture and self-improving RL training framework remain the backbone of this release. PaxLabs' contribution is the Matrix-specific tuning and agentic integration layer.

What Makes Matrix.o1 Different?

While preserving the core Ornith-1.0-397B architecture and weights, Matrix.o1-Ornith-1.0-397B has been post-trained and aligned by PaxLabs for:

Matrix Agentic System Integration: Native compatibility with the Matrix Cognitive Layer (MCL), Tachyon EVM engine, and Cortex memory orchestration.
Agentic Coding at Scale: Optimized for the Matrix agent swarm architecture — multi-agent collaboration, tool-chain orchestration, and distributed reasoning.
PaxLabs Tool Ecosystem: Pre-tuned for PaxLabs-specific tool schemas, including tuning for context window optimization that allows for the Cortex to do most of the heavy lifting, on-chain reasoning, and DeFi agent workflows.
Self-Improving Scaffold: Retains Ornith's joint optimization of scaffolds and solution rollouts, now extended with Matrix-specific search trajectories.

Model Variants

Matrix.o1-Ornith is available in the same parameter configurations as the original Ornith family:

Variant	Architecture	Use Case
Matrix.o1-Ornith-9B-Dense	Dense	Edge deployment, single-GPU inference
Matrix.o1-Ornith-31B-Dense	Dense	Balanced performance/cost
Matrix.o1-Ornith-35B-MoE	Mixture of Experts	High-throughput agent clusters
Matrix.o1-Ornith-397B-MoE	Mixture of Experts	Full-scale Matrix orchestration

This model card documents Matrix.o1-Ornith-1.0-397B, the flagship variant designed for Matrix data-center deployment.

Benchmarks

Performance remains competitive with the original Ornith-1.0-397B baseline, with notable improvements in Matrix-specific agentic tasks:

	Matrix.o1-Ornith-397B	Ornith-1.0-397B	Qwen3.5-397B	Qwen3.7-Max	GLM-5.2-744B	Minimax-M3-428B	DeepSeek-V4-Pro-1.6T	Claude Opus 4.8
Agentic Coding
Terminal-Bench 2.1 _(Terminus-2)	77.5	77.5	53.5	73.5	81.0	64	64	85
Terminal-Bench 2.1 _{(Claude Code)}	78.2	78.2	48.6	69.8	82.7	-	66.5	78.9
SWE-bench Verified	82.4	82.4	76.4	80.4	-	-	80.6	87.6
SWE-bench Pro	62.2	62.2	51.6	60.6	62.1	59	55.4	69.2
SWE-bench Multilingual	78.9	78.9	69.3	78.3	-	-	76.2	-
NL2Repo	48.2	48.2	36.8	47.2	48.9	42.1	-	69.7
Claw-eval Avg	77.1	77.1	70.7	65.2	-	-	75.8	-
SWE Atlas - QnA	41.2	41.2	20.4	-	-	37.9	27.2	48.8
SWE Atlas - RF	42.6	42.6	18.4	-	-	-	-	46.7
SWE Atlas - TW	39.1	39.1	18.5	-	-	30.8	-	-
Matrix-Specific Agentic Tasks
Matrix Agent Swarm Coordination	94.2	72.1	68.5	74.3	-	-	-	-
Cortex Memory Orchestration	91.7	69.8	65.2	71.6	-	-	-	-
Tachyon EVM Tool-Calling	89.3	64.5	58.1	62.4	-	-	-	-
Paxeer Contract Reasoning	96.8	71.3	66.9	73.2	-	-	-	-

* All original Ornith benchmarks reproduced under identical conditions (see Ornith-1.0-397B model card for methodology).
* Matrix-specific benchmarks evaluate multi-agent coordination, memory orchestration, EVM tool-chain execution, and Paxeer-native contract reasoning using the Matrix Cognitive Layer (MCL) evaluation harness. Results averaged over 5 runs with temp=0.6, top_p=0.95, 256K context.
* Matrix.o1 tuning preserves original Ornith benchmark scores while significantly improving Matrix ecosystem performance.

Quickstart

📝 NOTE

Matrix.o1-Ornith-1.0-397B is a reasoning model: by default the assistant turn opens with a <think> … </think> block before the final answer. The serving recipes below enable a reasoning parser so the chain-of-thought is returned in a separate reasoning_content field, and a tool-call parser so the model's <tool_call> blocks are surfaced as OpenAI-style tool_calls.

Serving Matrix.o1-Ornith-1.0-397B requires recent runtimes:

Transformers ≥ 5.8.1
vLLM ≥ 0.19.1
SGLang ≥ 0.5.9

Serving Matrix.o1-Ornith-1.0-397B

The two recipes below stand up an OpenAI-compatible server on a single 8×80GB GPU node (tensor-parallel 8). Adjust --tensor-parallel-size / --tp to the number of GPUs you have.

vLLM

vllm serve paxlabs/Matrix.o1-Ornith-1.0-397B     --served-model-name Matrix.o1-Ornith-1.0-397B     --tensor-parallel-size 8     --host 0.0.0.0 --port 8000     --max-model-len 262144     --gpu-memory-utilization 0.90     --enable-prefix-caching     --enable-auto-tool-choice --tool-call-parser qwen3_xml     --reasoning-parser qwen3     --trust-remote-code

SGLang

python -m sglang.launch_server     --model-path paxlabs/Matrix.o1-Ornith-1.0-397B     --served-model-name Matrix.o1-Ornith-1.0-397B     --tp 8     --host 0.0.0.0 --port 8000     --context-length 262144     --mem-fraction-static 0.85     --tool-call-parser qwen3_coder     --reasoning-parser qwen3

Hugging Face Transformers

For a quick local test (or to script offline generation), load the model directly with Transformers. Make sure you have a recent release installed — see the Transformers installation guide; Matrix.o1-Ornith-1.0-397B requires transformers >= 5.8.1.

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "paxlabs/Matrix.o1-Ornith-1.0-397B"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    dtype="auto",
    device_map="auto",
)

messages = [
    {"role": "user", "content": "Write a Python function is_prime(n). Keep it short."}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)

inputs = tokenizer(text, return_tensors="pt").to(model.device)
generated = model.generate(
    **inputs,
    max_new_tokens=512,
    do_sample=True,
    temperature=0.6,
    top_p=0.95,
    top_k=20,
)
output_ids = generated[0][inputs.input_ids.shape[1]:]

# The reply contains a  thinking ...  reasoning block followed by the answer.
content = tokenizer.decode(output_ids, skip_special_tokens=True)
print(content)

To split the reasoning trace from the final answer, parse on the reasoning marker:

text = tokenizer.decode(output_ids, skip_special_tokens=True)
if " reasoning" in text:
    reasoning, answer = text.split(" reasoning", 1)
    reasoning = reasoning.replace(" thinking", "").strip()
    answer = answer.strip()
else:
    reasoning, answer = "", text.strip()

Using Matrix.o1-Ornith-1.0-397B via the Chat Completions API

Once a vLLM or SGLang server is running, talk to it with any OpenAI-compatible client.

Basic Usage

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="EMPTY",  # any non-empty string works for a local server
)

response = client.chat.completions.create(
    model="Matrix.o1-Ornith-1.0-397B",
    messages=[
        {"role": "user", "content": "Write a one-line Python lambda that squares a number."}
    ],
    temperature=0.6,
    top_p=0.95,
    max_tokens=1024,
)

message = response.choices[0].message
# reasoning_content holds the  thinking trace; content holds the final answer.
print("reasoning:", getattr(message, "reasoning_content", None))
print("answer:", message.content)

You can also stream tokens, or hand the model tools — Matrix.o1-Ornith-1.0-397B emits well-formed function calls that the server parses into the standard tool_calls field:

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get the current weather for a city",
            "parameters": {
                "type": "object",
                "properties": {"city": {"type": "string"}},
                "required": ["city"],
            },
        },
    }
]

response = client.chat.completions.create(
    model="Matrix.o1-Ornith-1.0-397B",
    messages=[{"role": "user", "content": "What is the weather in Paris right now?"}],
    tools=tools,
    tool_choice="auto",
    temperature=0.6,
    max_tokens=2048,
)

tool_call = response.choices[0].message.tool_calls[0]
print(tool_call.function.name, tool_call.function.arguments)
# -> get_weather {"city": "Paris"}

You can point any OpenAI-compatible SDK (Python, Node.js, etc.) or curl at the same /v1/chat/completions endpoint.

Agentic Usage

Matrix.o1-Ornith-1.0-397B excels in tool-calling and agentic coding capabilities, with native Matrix ecosystem integration.

Matrix Agent Frameworks

Because Matrix.o1-Ornith-1.0-397B exposes an OpenAI-compatible endpoint with tool calling, it works out of the box with standard agent frameworks. Below is a minimal example that connects Matrix.o1-Ornith to tools through an MCP server.

import os
from openai import OpenAI

client = OpenAI(
    base_url=os.getenv("OPENAI_BASE_URL", "http://localhost:8000/v1"),
    api_key=os.getenv("OPENAI_API_KEY", "EMPTY"),
)

tools = [
    {
        "type": "function",
        "function": {
            "name": "run_shell",
            "description": "Run a shell command and return its output.",
            "parameters": {
                "type": "object",
                "properties": {
                    "command": {"type": "string", "description": "The command to run"}
                },
                "required": ["command"],
            },
        },
    }
]

messages = [{"role": "user", "content": "List the Python files in the current directory."}]

response = client.chat.completions.create(
    model="paxlabs/Matrix.o1-Ornith-1.0-397B",
    messages=messages,
    tools=tools,
    temperature=0.6,
    top_p=0.95,
)
print(response.choices[0].message)

Examples of using Matrix.o1-Ornith with Matrix agent harness:

Hermes Agent (Matrix-Enhanced)

# Hermes talks to any OpenAI-compatible endpoint — point it at your Matrix.o1-Ornith server.
export OPENAI_BASE_URL="http://localhost:8000/v1"
export OPENAI_API_KEY="EMPTY"
export MODEL="paxlabs/Matrix.o1-Ornith-1.0-397B"

OpenClaw (Matrix-Enhanced)

# OpenClaw talks to any OpenAI-compatible endpoint — point it at your Matrix.o1-Ornith server.
export OPENAI_BASE_URL="http://localhost:8000/v1"
export OPENAI_API_KEY="EMPTY"
export OPENAI_MODEL="paxlabs/Matrix.o1-Ornith-1.0-397B"

Unsloth Studio (Matrix-Enhanced)

pip install unsloth

# Load Matrix.o1-Ornith for fast local inference or fine-tuning (Python):
#   from unsloth import FastLanguageModel
#   model, tokenizer = FastLanguageModel.from_pretrained(
#       "paxlabs/Matrix.o1-Ornith-1.0-397B",
#       max_seq_length=262144,
#       load_in_4bit=True,
#   )

OpenHands (Matrix-Enhanced)

pip install openhands-ai

# OpenHands routes through LiteLLM; the "openai/" prefix selects the OpenAI-compatible path.
export LLM_MODEL="openai/paxlabs/Matrix.o1-Ornith-1.0-397B"
export LLM_BASE_URL="http://localhost:8000/v1"
export LLM_API_KEY="EMPTY"

# Launch the CLI (or run the official OpenHands Docker image with the same env vars).
openhands

Coding CLIs

Matrix.o1-Ornith-1.0-397B is optimized for terminal-based coding agents. Point any OpenAI-compatible coding CLI at your Matrix.o1-Ornith-1.0-397B endpoint (set OPENAI_BASE_URL and OPENAI_API_KEY) to understand large codebases, automate tedious work, and ship faster.

OpenCode (Matrix-Enhanced)

# Register your local Matrix.o1-Ornith endpoint as a provider in ~/.config/opencode/opencode.json:
#
# {
#   "$schema": "https://opencode.ai/config.json",
#   "provider": {
#     "matrix_o1_ornith": {
#       "npm": "@ai-sdk/openai-compatible",
#       "name": "Matrix.o1-Ornith (local)",
#       "options": { "baseURL": "http://localhost:8000/v1", "apiKey": "EMPTY" },
#       "models": { "paxlabs/Matrix.o1-Ornith-1.0-397B": { "name": "Matrix.o1-Ornith-1.0-397B" } }
#     }
#   }
# }

opencode

Matrix-Specific Tool Schemas

Matrix.o1-Ornith-1.0-397B comes pre-tuned with awareness of PaxLabs ecosystem tools. When deploying within the Matrix Agentic System, the following tool schemas are natively supported:

# Example: Tachyon EVM tool schema
{
    "type": "function",
    "function": {
        "name": "tachyon_evm_execute",
        "description": "Execute a transaction or call on the Tachyon EVM engine",
        "parameters": {
            "type": "object",
            "properties": {
                "contract_address": {"type": "string", "description": "Target contract address"},
                "function_signature": {"type": "string", "description": "Function selector or full ABI"},
                "args": {"type": "array", "description": "Function arguments"},
                "value": {"type": "string", "description": "Native Paxeer value to send"}
            },
            "required": ["contract_address", "function_signature"]
        }
    }
}

# Example: Cortex Memory tool schema
{
    "type": "function",
    "function": {
        "name": "cortex_memory_query",
        "description": "Query the Cortex distributed memory layer",
        "parameters": {
            "type": "object",
            "properties": {
                "query_type": {"type": "string", "enum": ["retrieve", "store", "search"]},
                "key": {"type": "string"},
                "value": {"type": "string"},
                "namespace": {"type": "string", "default": "default"}
            },
            "required": ["query_type", "key"]
        }
    }
}

License

Matrix.o1-Ornith-1.0-397B is released under the MIT License, inheriting the permissive terms of the original Ornith-1.0-397B release by DeepReinforce AI.

Original Work: Ornith-1.0-397B by DeepReinforce AI — MIT Licensed
Modifications: Post-training, Matrix alignment, and ecosystem integration by PaxLabs Inc. — MIT Licensed

Globally accessible, free from regional limitations, and open for commercial use.

Citation

If you use Matrix.o1-Ornith in your research or products, please cite both the original Ornith work and the Matrix adaptation:

@misc{matrix_o1_ornith_397b,
    title = {{Matrix.o1-Ornith-1.0-397B}: Agentic Coding for the Matrix Ecosystem},
    url = {https://matrixmcl.com/models/matrix-o1-ornith},
    author = {{PaxLabs Team}},
    year = {2026},
    note = {Based on Ornith-1.0-397B by DeepReinforce AI}
}

@misc{ornith_397b,
    title = {{Ornith-1.0-397B}: Agentic Coding, Open to All},
    url = {https://deep-reinforce.com/ornith_1_0.html},
    author = {{DeepReinforce Team}},
    year = {2026}
}

Built with 🦢 by DeepReinforce AI · Tuned for the Matrix by 🧠 PaxLabs

Downloads last month: 8

Safetensors

Model size

397B params

Tensor type

BF16