GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
morningstarxcdcode
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
morningstarxcdcode
Model Tree
google/gemma-4-31B-it
Input Modalities
Output Modalities
Supported Functionality
google/gemma-4-31B-it (31B parameters)q_proj and v_proj| Metric | Before | After |
|---|---|---|
| Quality | 2.0 | 5.9 (+195.0%) |
| Grade | E | C |
| Percentile | 0.1 | 7.2 |
| Win Rate | 46% | 54% |
python
from peft import PeftModelfrom transformers import AutoModelForCausalLM, AutoTokenizerbase_model = AutoModelForCausalLM.from_pretrained("google/gemma-4-31B-it",torch_dtype="bfloat16",device_map="auto")model = PeftModel.from_pretrained(base_model, "morningstarxcdcode/adaption-opus-100-translation-model")tokenizer = AutoTokenizer.from_pretrained("morningstarxcdcode/adaption-opus-100-translation-model")inputs = tokenizer("Translate to French: The weather is nice today.", return_tensors="pt").to(model.device)outputs = model.generate(**inputs, max_new_tokens=256)print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Sourav Rajak, Priyanshu Tomar, Roshan G, Vivek Rajput
Part of the AutoScientist Challenge — Healthcare, Finance, Language, Legal, and Marketing tracks.