GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
ceselder
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
ceselder
Model Tree
Qwen/Qwen3.6-27B
Input Modalities
Output Modalities
Supported Functionality
Part of the skip-lens project (multi-token NL activation lenses on Qwen3.6-27B).
This is the plain/normal future-lens. A LoRA adapter (r=64, α=16, rsLoRA) on Qwen/Qwen3.6-27B
trained by SFT on 485,000 (layer-62 activation, future-tokens) pairs. Given a single residual
activation injected at a fixed position (Karvonen norm-matched, char ㈜ / id 158983), it decodes
the text the model was about to produce next.
Sibling: skip-lens-qwen36-27b-futurelens-rl is the same lens after GRPO RL against a frozen
reconstruction reward. This repo is the SFT-only base future-lens.
Load as a PEFT adapter (PeftModel.from_pretrained) + inject a layer-62 activation via the Karvonen hook.