DeepSeek V4 Flash · Abliterated · GGUF CyberNeurova research — cyberneurova.ai Release: v2 (three direction multi turn aware ablation, 1338 prompt capture corpus) A permanently abliterated version of deepseek ai/DeepSeek V4 Flash (284 B FP8 MoE) packaged as GGUF for llama.cpp . The abliteration is baked into the weights at conversion time — no runtime hooks, no slowdown, no reliance on the inference framework supporting custom code paths. Status: experimental research artifact. Built with antirez's llama.cpp DeepSeek V4 Flash fork, which is itself experimental. Use at your own discretion. Audrey Tang DS4 re quants This fork also hosts re quantized GGUFs made from CyberNeurova's Q8 0 release for DS4 / pi ds4 testing. Motivation: provide antirez style q2 imatrix and q4 imatrix shapes that the stock ds4 loader can load directly, including an F16 token embedding, so the loader does not need the support q8 0 token embd workaround. Added files: Variant File Approx. size Notes : q2 imatrix cyberneurova DeepSeek V4 Flash abliterated IQ2XXS w2Q2K AProjQ8 SExpQ8 OutQ8 chat v2 imatrix.gguf 81 GB IQ2 XXS gate/up routed experts, Q2 K routed down experts, Q8 attention/shared/output paths q2 imat…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy