Released by @dealignai · Compute generously provided by @jordanschenck

DeepSeek-V4.1-Flash-1017 — ABLITERATED (FP8)

👉 This is an alias — the actual model weights live at dealignai/DeepSeek-V4.1-Flash-1017-UNCENSORED-FP8.

---

Please use the canonical repo: https://huggingface.co/dealignai/DeepSeek-V4.1-Flash-1017-UNCENSORED-FP8

That repo contains the model weights, tokenizer, inference code (inference/), prompt-encoding module (encoding/), benchmarks, run instructions, and the gated-access form.

This name exists only as a canonical alternate reference for users who search for "abliterated" rather than "uncensored" — same model, same ship date, same release owner.

[!IMPORTANT] Gating is temporary — it will be lifted (public download) once the full HarmBench-320 and MMLU-14k evaluations complete and the final numbers land on the UNCENSORED card. Until then, access is approved case-by-case.

[!CAUTION]

Attribution is required

If you redistribute, quantize (GGUF / MLX / EXL / AWQ / anything), fine-tune, serve, benchmark, or otherwise build on this model: please prominently credit dealignai as the source and link back to the UNCENSORED repository in your model card.

Credits

• Base model by DeepSeek-AI. Reference inference/ engine and encoding/ module by DeepSeek-AI (MIT).
• Release and model card by @dealignai.
• H200 compute for building this release was generously provided by @jordanschenck.