GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
aziz9788
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
aziz9788
Model Tree
unsloth/Qwen3.5-4B
Input Modalities
Output Modalities
Supported Functionality
Fully merged Qwen3.5-4B weights (Unsloth base + LoRA + identity SFT). Load as a standalone Transformers / vLLM model — no adapter required.
Identity recipe: replay-r32-e3-5pct (transformers+peft bf16 LoRA, no Unsloth), thinking off. Base: /workspace/Saudi-LLM/RL/DPO/artifacts/qwen35_direct_merged. Source adapter: aziz9788/qwen35-saudidraft-full-final-adapter. Run: saudidraft_direct_20260822_200129. Uploaded after identity merge, before the post-identity suite. Pre-identity suite: /workspace/Saudi-LLM/Eval/outputs/full_suite_no_reasoning_saudidraft_direct_20260822_200129_pre_identity.