An uncensored GGUF build of GLM-5.2 that applies weight “abliteration” to remove refusal filters and produce a locally runnable text-generation model; includes quantization conversions and shard-merge instructions, intended for experimental research rather than production use.
Thinking-off fine-tune for coding-agent workflows that prioritizes fast next-step decisions, lower token usage and stable multi-turn tool calling. Highlights: MoE 35B base, MTP speculative decoding, SWE-bench 62.4% (300 cases). Best for local agent loops and automated debug cycles; requires disciplined harnessing and schema consistency.
Uncensored local builds are often created for red‑teaming, research, or offline experimentation. This release demonstrates a proof‑of‑concept approach that directly modifies GGUF weights of GLM‑5.2 (“abliteration”) to suppress refusal behaviors, trading safety filtering for broader output freedom.
Great fit if you need a locally runnable GLM‑5.2 variant for controlled research, red‑teaming, or experiments where standard refusal behavior is a blocker and you can assume manual output review. The model is useful to study the effect of weight‑level interventions on safety and generation behavior.
Look elsewhere if you require a production‑grade, safety‑hardened model or if legal/ethical constraints forbid generating uncensored content. Trade‑offs include increased risk of sensitive or inappropriate outputs, potential legal/ethical exposure, greater responsibility for monitoring, and the need for significant local compute and manual merging steps. Use in public or underage contexts is not recommended.
