license: apache-2.0
tags:
- gguf
- qwen3.5
- moe
- vision
- uncensored
- tool-calling
- cypher-studio
base_model: HauhauCS/Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive
Cypher Qwen3.5 35B-A3B Uncensored (GGUF + Vision)
Miroir souverain Cypher Studio (cypher-opencode) optimise pour Kaggle Dual-T4 (30 Go VRAM) sous llama-server --jinja.
- Architecture : MoE 35B total / 3B actifs + Gated DeltaNet 3:1 (KV cache ultra-compact)
- Quantification :
Q4_K_Mimatrix (19.72 GiB) +mmproj-f16(0.84 GiB, vision native) - Capacites : Thinking (
<think>), Tool Calling natif, Vision multimodale, contexte 64K-256K - Source amont :
HauhauCS/Qwen3.5-35B-A3B-Uncensored-HauhauCS-Aggressive