Instructions to use open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10
- SGLang
How to use open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10 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 "open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10" \ --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": "open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10", "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 "open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10" \ --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": "open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10 with Docker Model Runner:
docker model run hf.co/open-athena/Grug-67B-A2B-Merge-Campaign2-DARE-TIES-02-2026.10.10
Grug 67B A2B — Campaign 2 DARE-TIES-02
Research artifact. This model has not been properly tested or evaluated and is not necessarily secure. Use at your own risk in production settings.
This is the dare-ties-02 checkpoint from the second Snowball/Grug model-merging campaign. It transfers specialist updates into the Base checkpoint while preserving 93.96% of Base scalar weights bit-for-bit. The measured changed fraction is 6.037362616948686%, within the campaign's 25% cap.
| Fixed-policy benchmark | Score (%) |
|---|---|
| NUPA, fixed 200-task subset | 0.00 |
| BFCL-Parity/Pi | 45.53 |
| Sum | 45.53 |
NUPA scored all 200 tasks; BFCL scored all 123 trials. The NUPA zero received the required ten-trace review. These are development benchmark results, not held-out evidence. Non-target capability retention was not evaluated for this checkpoint.
The merge applies DARE-TIES within the fixed R25 routed-expert support (1,705 expert blocks). DARE keep density is 0.20, TIES trimming density is 0.20, the NUPA/BFCL delta coefficients are 0.50/0.50, the scale is 1.00, and the random seed is 0. Outside the selected support, weights are copied from Base unchanged. Tokenizer, chat template, model configuration, tensor names, dtypes and shard layout are unchanged from Base.
Parent checkpoints are pinned as follows:
- Base: open-athena/Grug-67B-A2B-Antidoom-RLVR1-Step12-2026.10.02 at
50fe2f488aee7730bbccc721024b35229075602e. - NUPA: open-athena/Snowball-67B-A2B-10T-Mixed-RLVR-Sync-Step92 at
fdc3c3c5fff489d3fb04dcbc2387c179b1bf3fab. - BFCL: open-athena/Grug-67B-A2B-Antidoom-RLVR1-Reference-KL-Step20-2026.10.04 at
fd07ef0f2b78a83099a0042e59c00de986e7b647.
Merge provenance and file hashes contain the exact support, merge parameters and SHA256 hashes of all 48 source files. All 48 source files are published unchanged; this model card, the license and merge provenance are publication additions. The inherited .cache_metadata.json and .marinskyrl-model-manifest.json describe Base's original cache identity and remain unchanged for source fidelity; use merge-provenance.json to verify this merged checkpoint's files.
Licensed under OpenMDW 1.1.
We release our tools to the public with the intent that they are used safely, ethically, and responsibly. They are meant to contribute to the public's understanding of science, and aid in its advancement. By using this Marin model, you agree to comply with all applicable law.
- Downloads last month
- 93