Instructions to use Kronumos/Kronumos-Metis with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Kronumos/Kronumos-Metis with Transformers:

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

pipe = pipeline("text-generation", model="Kronumos/Kronumos-Metis", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Kronumos/Kronumos-Metis", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("Kronumos/Kronumos-Metis", trust_remote_code=True, device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Kronumos/Kronumos-Metis with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Kronumos/Kronumos-Metis

SGLang

How to use Kronumos/Kronumos-Metis 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 "Kronumos/Kronumos-Metis" \
    --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": "Kronumos/Kronumos-Metis",
		"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 "Kronumos/Kronumos-Metis" \
        --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": "Kronumos/Kronumos-Metis",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Kronumos/Kronumos-Metis with Docker Model Runner:
```
docker model run hf.co/Kronumos/Kronumos-Metis
```

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Kronumos-Metis is an enterprise closed-weight cybernetic model for autonomous program repair. Submit your organization credentials to request evaluation access.

Kronumos-Metis (236B MoE)

Enterprise Workhorse Autonomous Program Repair • DeepSeek MLA Cybernetic Architecture

Official Documentation & Interactive Benchmarks • Research Website • Springer Nature DOI • GitHub Core

High-throughput enterprise mixture-of-experts model weights for multi-file repository repair and autonomous software engineering.

Kronumos Metis

Tactical Cybernetic Engine • 236B Mixture-of-Experts (21B Active)

Abstract & Positioning

Kronumos-Metis is the enterprise workhorse foundation of the Kronumos autonomous program repair family, engineered by Tokenectomy Labs. Named after the Greek goddess of tactical ingenuity (Metis, Μῆτις), it delivers dense-grade reasoning with the sub-millisecond execution throughput of a 21B sparse model.

Paired natively with the Tokenectomy Sub-Cortex (Rust C-ABI, 5µs), Metis eliminates conversational bloat, validates Tree-sitter AST syntax before patch emission, and achieves production-grade patch synthesis across Python, Rust, C, C++, and TypeScript.

Key Architectural Specifications

Total Parameters: 236 Billion
Active Parameters per Token: 21 Billion (Sparse MoE)
Attention Mechanism: Multi-Head Latent Attention (MLA) with 5x-8x compressed KV cache footprint
Context Window: 128,000 tokens
Architecture Family: DeepSeek MoE (Unified lineage with Kronumos-Aion)
Runtime Orchestrator: Tokenectomy Sub-Cortex (libtokenectomy_subcortex.so, 5µs latency)
Target Workloads: Enterprise CI/CD autonomous ticket remediation, multi-file architectural refactoring, production incident triage

Enterprise Gated Access

Kronumos-Metis is distributed under the Tokenectomy Enterprise Dual License (TDL 1.0) for managed cloud API and air-gapped sovereign on-premise installations.

To request evaluation access:

Complete the Hugging Face access request form above.
Inquiries and enterprise pilot requests may also be directed to: daffa@kronumos.com.

Academic & Research Provenance

@article{daffa2026kronumos,
  title={Neurosymbolic Cybernetics for Deterministic Autonomous Program Repair},
  author={Daffa, Muhammad Naufal},
  journal={Springer Nature Preprint},
  year={2026},
  doi={10.21203/rs.3.rs-11205335/v1}
}

Downloads last month: 3

Safetensors

Model size

236B params

Tensor type

BF16

Dataset used to train Kronumos/Kronumos-Metis

Collection including Kronumos/Kronumos-Metis

Kronumos-Metis

Collection

1 item • Updated 6 days ago