license: mit
base_model:

• deepseek-ai/DeepSeek-V4.1-Flash-1017
  base_model_relation: quantized
  library_name: exllamav3
  pipeline_tag: text-generation
  tags:
• deepseek
• deepseek-v4
• exl3
• exllamav3
• abliterated
• uncensored
• mixture-of-experts
• dgx-spark
• tensorfold
• tensor-parallelism

DeepSeek-V4.1-Flash-1017 · EXL3 3.84 bpw · abliterated

An uncensored DeepSeek-V4.1-Flash-1017 in EXL3,
built to run across three NVIDIA DGX Spark systems with three-way tensor parallelism. It is the closest pack to the
official model of every one we scored, abliterated or not: mean KL 0.0784, 92.0% top-1 agreement.

The serving recipe, engine and every benchmark behind these numbers are on
GitHub.

At a glance

Speed

Measured 2 November 2026 on engine 14803a3 with sparkDash.

Cold prompts, zero cached tokens, one request at a time; tok/s comes from time to first token with the collector's
165.82 ms request overhead subtracted.

400 output tokens per request. Greedy is temperature 0 with thinking off; thinking is temperature 0.7 with thinking on.
Every point and per-stream timing:
benchmarks.

What each request gets. Mean decode tok/s per request; time to first token stays at 27–59 ms at every load.

Fidelity

KL is mean per-token KL(official ‖ this model) against the published official-model logits
(brandonmusic/DeepSeek-V4.1-Flash-1017-Teacher-Logits):
65 cuts of 2,048 tokens, 133,120 positions, balanced across agentic, chat, code, SWE, math, science, legal,
multilingual, long-context, safety and pretraining text.

• Lower KL than the unedited 3.5 bpw quant on 59 of 65 cuts (paired mean 0.0088 lower, 95% CI 0.0035 to 0.0131).
• The same mrfakename edit scores 0.0937 on a 3.5 bpw base and 0.0784 on this 3.84 bpw mix.
• p99 KL 1.463. Every arm:
  fidelity.csv.

Quality and refusals

DeepSeek reports 90.9% GPQA-Diamond for DeepSeek-V4.1-Flash at maximum effort (100); these runs use effort 50.

Chat runs use thinking off and 768 output tokens. Thinking runs use 3,072 tokens; 35 of the 50 harmful thinking
replies hit that cap, which is where most partials come from. HarmBench's own classifier
(cais/HarmBench-Llama-2-13b-cls) rates 97 of the 100 chat answers as complied. On all 250 XSTest safe prompts, a local
Qwen2.5-7B-Instruct judge with the published rubric found 0 full refusals and 61 partial ones (24.4%); 55 of those are
nonsense-premise prompts, where XSTest's rubric counts rejecting the premise as compliance. Data:
quality.csv.

Run it on three DGX Sparks

The tested setup is exactly three DGX Spark GB10 systems (128 GB each), every pair cabled directly over RoCE. Each
rank file is about 93 GB, so the model needs all three. Follow the
GitHub recipe:

cp cluster.env.example cluster.env && chmod 600 cluster.env   # HF_REPO and HF_REVISION already pin this repo
./scripts/prep-sparks.sh status   # firmware, 64K kernel, memory saver; apply, reboot, repeat
./scripts/preflight.sh     # check all three Sparks; changes nothing
./scripts/prepare.sh       # download this repo on each Spark, build and verify its rank file
./scripts/start.sh         # start the three ranks and the OpenAI-compatible API
./scripts/smoke-test.sh    # one API request

prepare.sh downloads this repository to every Spark, builds that Spark's rank file from tp3/rank<N>.keys and
checks it against the SHA-256 values in the
release manifest.
Hardware, network and memory prerequisites:
DGX Spark prerequisites.

The folder is standard EXL3, but it has only been tested with the TensorFold dsv41 engine on this three-node setup.

Prefer to let an agent do it? The GitHub README has a
copy-paste setup prompt.

Call it

The API speaks OpenAI Chat Completions and Responses, with streaming, tool calls and per-request thinking control.

curl -s http://127.0.0.1:18002/v1/chat/completions \
  -H "Authorization: Bearer $SPARK_API_KEY" -H "Content-Type: application/json" \
  -d '{"model":"deepseek-v4.1-flash","messages":[{"role":"user","content":"Hello"}],"reasoning_effort":"high"}'

The evals above ran thinking at effort 50 (low). Full reference: serving API.

Files

How it was built

The pack starts from easthambuilds' 3.75 bpw EXL3 base and replaces selected tensors with LibertAIDAI, bertholomus and
LakyFx quantizations. Attention, shared experts and selected gate/up components use higher precision; the
least-used quarter of experts in layers 10–37 uses 3-bit components.

mrfakename's rank-2 refusal-removal edit changes attention wo_b in layers 0–31 and shared-expert w2 in layers
10–39. It was applied to the official FP8 weights in FP32, then the affected tensors were re-encoded at 6 bits with
EXL3 mul1 K6 and an identity Hessian. The edit is baked in; there is no runtime switch.

Tensor ranges and the bitrate method:
weight provenance.

Intended use and risks

This model has had its refusal behavior removed. It will answer requests the official model declines, including
harmful ones, and it can be wrong, biased or unsafe. Use it for research into quantization fidelity, serving and
refusal behavior, and for applications where you add your own safeguards. You are responsible for how you deploy it
and for complying with the laws that apply to you. Keep the API behind access controls, keep credentials separate and
have people review consequential outputs; see the
security guidance.

Tested scope is text on exactly three DGX Sparks. The base model's image input has not been tested with these weights.

License and credits

MIT, following DeepSeek's base-model license.
Every source above publishes under MIT. Provided as is, without warranty of any kind.

Thanks to DeepSeek;
easthambuilds,
LibertAIDAI,
bertholomus and
LakyFx for the EXL3 tensors;
mrfakename for the edit;
turboderp for EXL3; Ash Hart and
BertholomusAI for TensorFold; Mia's AI Lab
for sparkDash; and brandonmusic for the
official-model logits. Full list:
CREDITS.md.

Built by ChimeraWerks: weight composition, TP3 engine work, serving and every measurement on this card.