Views
No views yet
W4A16 (INT4 weights, 16-bit activations)| Hyperparameter | Value |
|---|---|
| Total parameters | ~398B |
| Active parameters per token | ~13B |
| Experts | 256 (1 shared) |
| Active experts | 4 |
| Routing strategy | 4-of-256 (1.56% sparsity) |
| Dense layers | 6 |
| Pretraining context length | 8,192 |
| Context length after extension | 512k |
| Architecture | Sparse MoE (AfmoeForCausalLM) |
| Benchmark | Llama 4 Maverick | Trinity-Large Preview |
|---|---|---|
| MMLU | 85.5 | 87.2 |
| MMLU-Pro | 80.5 | 75.2 |
| GPQA-Diamond | 69.8 | 63.3 |
| AIME 2025 | 19.3 | 24.0 |
main transformers branch or pass trust_remote_code=True with a released version.1from transformers import AutoTokenizer, AutoModelForCausalLM
2import torch
3
4model_id = "arcee-ai/Trinity-Large-Preview-W4A16"
5tokenizer = AutoTokenizer.from_pretrained(model_id)
6model = AutoModelForCausalLM.from_pretrained(
7 model_id,
8 device_map="auto",
9 trust_remote_code=True
10)
11
12messages = [
13 {"role": "user", "content": "Who are you?"},
14]
15
16input_ids = tokenizer.apply_chat_template(
17 messages,
18 add_generation_prompt=True,
19 return_tensors="pt"
20).to(model.device)
21
22outputs = model.generate(
23 input_ids,
24 max_new_tokens=256,
25 do_sample=True,
26 temperature=0.8,
27 top_k=50,
28 top_p=0.8
29)
30
31response = tokenizer.decode(outputs[0], skip_special_tokens=True)
32print(response)1vllm serve arcee-ai/Trinity-Large-Preview-W4A16 \
2 --enable-auto-tool-choice \
3 --tool-call-parser hermes1curl -X POST "https://openrouter.ai/v1/chat/completions" \
2 -H "Authorization: Bearer $OPENROUTER_API_KEY" \
3 -H "Content-Type: application/json" \
4 -d '{
5 "model": "arcee-ai/trinity-large-preview",
6 "messages": [
7 {
8 "role": "user",
9 "content": "What are some fun things to do in New York?"
10 }
11 ]
12 }'1@misc{singh2026arceetrinity,
2 title = {Arcee Trinity Large Technical Report},
3 author = {Varun Singh and Lucas Krauss and Sami Jaghouar and Matej Sirovatka and Charles Goddard and Fares Obied and Jack Min Ong and Jannik Straube and Fern and Aria Harley and Conner Stewart and Colin Kealty and Maziyar Panahi and Simon Kirsten and Anushka Deshpande and Anneketh Vij and Arthur Bresnu and Pranav Veldurthi and Raghav Ravishankar and Hardik Bishnoi and DatologyAI Team and Arcee AI Team and Prime Intellect Team and Mark McQuade and Johannes Hagemann and Lucas Atkins},
4 year = {2026},
5 eprint = {2602.17004},
6 archivePrefix= {arXiv},
7 primaryClass = {cs.LG},
8 doi = {10.48550/arXiv.2602.17004},
9 url = {https://arxiv.org/abs/2602.17004}
10}