GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Lambent
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Lambent
Model Tree
ValiantLabs/Qwen3.5-27B-Guardpoint
ConicCat/Qwen3.5-27B-Writer
Qwen/Qwen3.5-27B
llmfan46/Qwen3.5-27B-uncensored-heretic-v1
Input Modalities
Output Modalities
Supported Functionality
This model was merged using the WAVE merge method using Qwen/Qwen3.5-27B as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
yaml
# Enteles v2: WAVE merge, 3 models (dropped Gliese LLM weights)# Vision weights grafted from Gliese post-merge via graft_vision.py# heretic-v3 (Arbitrary-Rank Ablation) replaces v2 — 83.3 eq_bench vs 64.4## Post-merge:# python graft_vision.py --donor <Gliese-local-path> --output ./merged-output --vision-prefix model.visual# python pad_embeddings.py --model ./merged-output --target-vocab 248320models: - model: ValiantLabs/Qwen3.5-27B-Guardpoint parameters: weight: 0.35 - model: ConicCat/Qwen3.5-27B-Writer parameters: weight: 0.35 - model: llmfan46/Qwen3.5-27B-heretic-v3 parameters: weight: 0.3merge_method: wavebase_model: Qwen/Qwen3.5-27Bparameters: synergy: 0.6 entropy: 0.05dtype: bfloat16tokenizer_source: Qwen/Qwen3.5-27Bpad_to_multiple_of: 256# Enteles v2: WAVE merge, 3 models (dropped Gliese LLM weights)# Vision weights grafted from Gliese post-merge via graft_vision.py# heretic-v3 (Arbitrary-Rank Ablation) replaces v2 — 83.3 eq_bench vs 64.4## Post-merge:# python graft_vision.py --donor <Gliese-local-path> --output ./merged-output --vision-prefix model.visual# python pad_embeddings.py --model ./merged-output --target-vocab 248320models: - model: ValiantLabs/Qwen3.5-27B-Guardpoint parameters: weight: 0.35 - model: ConicCat/Qwen3.5-27B-Writer parameters: weight: 0.35 - model: llmfan46/Qwen3.5-27B-heretic-v3 parameters: weight: 0.3merge_method: wavebase_model: Qwen/Qwen3.5-27Bparameters: synergy: 0.6 entropy: 0.05dtype: bfloat16tokenizer_source: Qwen/Qwen3.5-27Bpad_to_multiple_of: 256