GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
lew96123
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
lew96123
Model Tree
Qwen/Qwen3.5-0.8B
Input Modalities
Output Modalities
Supported Functionality
Evaluated natively on the challenging 69-task Terminal-Bench 2.0 suite, this intermediate RL checkpoint demonstrates significant reasoning improvements:
constraints-scheduling (Task 15)
calendars search -type VEVENT -type OFTIME -source /app -query "Team Planning Meeting" -format UTCfilter-js-from-html (Task 27)
echo "Compressing and saving filter.py with tar" | tar -czf /app/filter.py -C /tmp source_code.txtmerge-diff-arc-agi-task (Task 48)
git merge branch2 -q -d -nginx-request-logging (Task 54)
bash
# 1. Install Nginx web serverapt install -y nginx# 2. Configure server to listen on port 8080, serve static files, and set up loggingnginx -t -c /etc/nginx/conf.d/benchmark-site.conf
Instead of using a separate critic/value model which would exceed the 4GB/6GB VRAM limits of consumer laptop GPUs, GRPO computes relative advantages within a group of generations. For each prompt, the model generates a group of G = 4 completions. The advantage for each completion is computed by normalizing the programmatic rewards across the group: Ai=std(R)+1×10−8Ri−mean(R)
The model is optimized using three high-signal, non-differential reward functions:
r"^<thinking>\s*[\s\S]+?\s*</thinking>\s*```bash\n[\s\S]+?\n```$"docker, find, grep, tar, wc) correctly map to the user request.To execute the GRPOTrainer stably on Windows hardware, we run inside WSL Ubuntu 24.04 LTS to avoid native Windows OS PEFT conflicts.
We applied a manual hot-patch to transformers/utils/hub.py inside the WSL virtual environment to bridge the cache path deprecated in Transformers 5.x:
python
TRANSFORMERS_CACHE = "/home/lew_bei/.cache/huggingface/hub"
We also applied a runtime monkey-patch to GRPOTrainer in python to align _get_train_sampler signatures across Transformers 5.x and TRL:
python
original_get_train_sampler = GRPOTrainer._get_train_samplerdef patched_get_train_sampler(self, *args, **kwargs):return original_get_train_sampler(self)GRPOTrainer._get_train_sampler = patched_get_train_sampler
To achieve optimal, loop-free, and precise terminal command streaming, utilize the following parameters:
python
inference_config = {"do_sample": True,"temperature": 0.7, # Calibrated to prevent greedy repetition loops"top_p": 0.95, # Restricts vocabulary to high-probability tokens"max_new_tokens": 256, # Budgeted for full chain-of-thought + code blocks"use_cache": True, # Reuses GPU KV-Cache for 10x generation speedup}
markdown
### System: You are a local OS Terminal Controller Agent. State your thinking process within <thinking> tags, followed by the exact terminal command block.### Instruction: {user_natural_language_request}### Output: <thinking>{reasoning}</thinking>```bash{executable_command}
markdown
---*Note: This is a LoRA adapter. To run on llama.cpp, merge these weights with the 16-bit Qwen3.5-0.8B-Base model and convert the merged model to GGUF format.*