GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
amao0o0
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
amao0o0
Model Tree
deepseek-ai/DeepSeek-R1-Distill-Llama-8B
Input Modalities
Output Modalities
Supported Functionality
Accuracy (pass@1) and mean generated length averaged over AMC23, AIME24, MATH500 and GPQA-Diamond. AES is the Accuracy–Efficiency Score (α=1, β=3, γ=5).
| Model | Accuracy | Length | AES |
|---|---|---|---|
| DeepSeek-R1-Distill-Llama-8B | 47.6 | 9.4k | — |
| InfoDensity-DeepSeek-R1-Distill-Llama-8B | 58.4 | 6.5k | +0.99 |
Per benchmark (accuracy % / mean response tokens):
| Model | AMC23 | AIME24 | MATH500 | GPQA-D |
|---|---|---|---|---|
| DeepSeek-R1-Distill-Llama-8B | 65.0 / 8.5k | 23.3 / 13.3k | 78.8 / 5.2k | 23.2 / 10.7k |
| InfoDensity | 80.0 / 5.4k | 40.0 / 10.0k | 82.8 / 3.2k | 30.8 / 7.6k |
Both rows are scored with the released evaluation code, so a run of it reproduces these numbers.
Greedy decoding, pass@1, generation capped at 16384 tokens, with the prompt
markdown
<question>Please reason step by step, and put your final answer within \boxed{}
The evaluation scripts that produce these numbers are in the GitHub repository.
python
from transformers import AutoModelForCausalLM, AutoTokenizermodel_id = "amao0o0/InfoDensity-DeepSeek-R1-Distill-Llama-8B"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="bfloat16", device_map="auto")question = "What is the smallest positive integer n such that n^2 + n + 41 is not prime?"messages = [{"role": "user","content": f"{question}\n\nPlease reason step by step, and put your final answer within \\boxed{{}}"}]prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)out = model.generate(**tokenizer(prompt, return_tensors="pt").to(model.device), max_new_tokens=16384)print(tokenizer.decode(out[0], skip_special_tokens=True))
With vLLM:
python
from vllm import LLM, SamplingParamsllm = LLM(model="amao0o0/InfoDensity-DeepSeek-R1-Distill-Llama-8B")print(llm.generate([prompt], SamplingParams(temperature=0, max_tokens=16384))[0].outputs[0].text)
Released under mit, inherited from the base model deepseek-ai/DeepSeek-R1-Distill-Llama-8B.
bibtex
@inproceedings{wei2026infodensity,title = {InfoDensity: Rewarding Information-Dense Traces for Efficient Reasoning},author = {Wei, Chengwei and Kim, Jung-jae and Zhang, Longyin and Chen, Shengkai and Chen, Nancy F.},booktitle = {Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing},year = {2026}}