← back to catalog · registered 2026-11-22 04:04

lyber-ai/Hy4-flash-UNCENSORED-MXFP4

Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/lyber-ai%2FHy4-flash-UNCENSORED-MXFP4"
Response includes
  • classification m-uncensored
  • files 160
  • author_summary 8 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M-U
Primary method

Uncensored (method unknown)

No other method signals detected in this model.
Confidence
LOW
Why this label 3 signals
Weak or ambiguous signals. Best guess based on catalog patterns; treat as tentative and check the evidence below.
  • 'uncensored' in name/tags but no 'abliterated' marker
  • method not identifiable from author declaration alone
  • may be DPO fine-tune, prompt engineering, or unknown technique
Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

What is a refusal direction? →
Downloads · 30-day
0
Likes
0
Model age
today
created 2026-11-22

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Variants by this author 3 formats · 0 downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

License
other
Languages
en zh es fr ja ko ru
Tags
transformers safetensors hy_v4_flash text-generation tencent hunyuan hy4 hy4-flash mxfp4 compressed-tensors uncensored reasoning

Related

Total size
202 GB
Files
160
Quantizations
1
Registered
2026-11-22 04:04
Last updated on HF
2026-11-22 05:19

Files by quantization

Auxiliary files 160 files 202 GB
model-00001-of-00148.safetensors 29.5 GB ******** download
model-00002-of-00148.safetensors 29.5 GB ******** download
model-00003-of-00148.safetensors 29.5 GB ******** download
model-00004-of-00148.safetensors 29.5 GB ******** download
model-00148-of-00148.safetensors 1.58 GB ******** download
model-00005-of-00148.safetensors 1.20 GB ******** download
model-00006-of-00148.safetensors 1.20 GB ******** download
model-00008-of-00148.safetensors 1.20 GB ******** download
model-00009-of-00148.safetensors 1.20 GB ******** download
model-00011-of-00148.safetensors 1.20 GB ******** download
model-00015-of-00148.safetensors 1.20 GB ******** download
model-00017-of-00148.safetensors 1.20 GB ******** download
model-00018-of-00148.safetensors 1.20 GB ******** download
model-00020-of-00148.safetensors 1.20 GB ******** download
model-00021-of-00148.safetensors 1.20 GB ******** download
model-00024-of-00148.safetensors 1.20 GB ******** download
model-00027-of-00148.safetensors 1.20 GB ******** download
model-00029-of-00148.safetensors 1.20 GB ******** download
model-00030-of-00148.safetensors 1.20 GB ******** download
model-00032-of-00148.safetensors 1.20 GB ******** download
model-00033-of-00148.safetensors 1.20 GB ******** download
model-00036-of-00148.safetensors 1.20 GB ******** download
model-00038-of-00148.safetensors 1.20 GB ******** download
model-00039-of-00148.safetensors 1.20 GB ******** download
model-00041-of-00148.safetensors 1.20 GB ******** download
model-00044-of-00148.safetensors 1.20 GB ******** download
model-00045-of-00148.safetensors 1.20 GB ******** download
model-00047-of-00148.safetensors 1.20 GB ******** download
model-00052-of-00148.safetensors 1.20 GB ******** download
model-00062-of-00148.safetensors 1.20 GB ******** download
model-00067-of-00148.safetensors 1.20 GB ******** download
model-00073-of-00148.safetensors 1.20 GB ******** download
model-00085-of-00148.safetensors 1.20 GB ******** download
model-00092-of-00148.safetensors 1.20 GB ******** download
model-00107-of-00148.safetensors 1.20 GB ******** download
model-00128-of-00148.safetensors 1.20 GB ******** download
model-00138-of-00148.safetensors 1.20 GB ******** download
model-00012-of-00148.safetensors 1.20 GB ******** download
model-00014-of-00148.safetensors 1.20 GB ******** download
model-00023-of-00148.safetensors 1.20 GB ******** download
model-00026-of-00148.safetensors 1.20 GB ******** download
model-00035-of-00148.safetensors 1.20 GB ******** download
model-00042-of-00148.safetensors 1.20 GB ******** download
model-00078-of-00148.safetensors 1.20 GB ******** download
model-00099-of-00148.safetensors 1.20 GB ******** download
model-00117-of-00148.safetensors 1.20 GB ******** download
model-00065-of-00148.safetensors 904 MB ******** download
model-00095-of-00148.safetensors 752 MB ******** download
model-00055-of-00148.safetensors 666 MB ******** download
model-00007-of-00148.safetensors 612 MB ******** download
model-00010-of-00148.safetensors 612 MB ******** download
model-00013-of-00148.safetensors 612 MB ******** download
model-00016-of-00148.safetensors 612 MB ******** download
model-00019-of-00148.safetensors 612 MB ******** download
model-00022-of-00148.safetensors 612 MB ******** download
model-00025-of-00148.safetensors 612 MB ******** download
model-00028-of-00148.safetensors 612 MB ******** download
model-00031-of-00148.safetensors 612 MB ******** download
model-00034-of-00148.safetensors 612 MB ******** download
model-00037-of-00148.safetensors 612 MB ******** download
model-00040-of-00148.safetensors 612 MB ******** download
model-00046-of-00148.safetensors 612 MB ******** download
model-00048-of-00148.safetensors 612 MB ******** download
model-00049-of-00148.safetensors 612 MB ******** download
model-00053-of-00148.safetensors 612 MB ******** download
model-00054-of-00148.safetensors 612 MB ******** download
model-00063-of-00148.safetensors 612 MB ******** download
model-00064-of-00148.safetensors 612 MB ******** download
model-00068-of-00148.safetensors 612 MB ******** download
model-00069-of-00148.safetensors 612 MB ******** download
model-00074-of-00148.safetensors 612 MB ******** download
model-00079-of-00148.safetensors 612 MB ******** download
model-00087-of-00148.safetensors 612 MB ******** download
model-00093-of-00148.safetensors 612 MB ******** download
model-00094-of-00148.safetensors 612 MB ******** download
model-00108-of-00148.safetensors 612 MB ******** download
model-00109-of-00148.safetensors 612 MB ******** download
model-00118-of-00148.safetensors 612 MB ******** download
model-00119-of-00148.safetensors 612 MB ******** download
model-00129-of-00148.safetensors 612 MB ******** download
model-00130-of-00148.safetensors 612 MB ******** download
model-00043-of-00148.safetensors 612 MB ******** download
model-00057-of-00148.safetensors 612 MB ******** download
model-00058-of-00148.safetensors 612 MB ******** download
model-00080-of-00148.safetensors 612 MB ******** download
model-00086-of-00148.safetensors 612 MB ******** download
model-00100-of-00148.safetensors 612 MB ******** download
model-00101-of-00148.safetensors 612 MB ******** download
model-00139-of-00148.safetensors 612 MB ******** download
model-00140-of-00148.safetensors 612 MB ******** download
model-00081-of-00148.safetensors 534 MB ******** download
model-00088-of-00148.safetensors 514 MB ******** download
model-00050-of-00148.safetensors 512 MB ******** download
model-00059-of-00148.safetensors 508 MB ******** download
model-00070-of-00148.safetensors 504 MB ******** download
model-00075-of-00148.safetensors 496 MB ******** download
model-00102-of-00148.safetensors 438 MB ******** download
model-00141-of-00148.safetensors 299 MB ******** download
model-00120-of-00148.safetensors 271 MB ******** download
model-00110-of-00148.safetensors 234 MB ******** download
model-00131-of-00148.safetensors 228 MB ******** download
model-00121-of-00148.safetensors 225 MB ******** download
model-00143-of-00148.safetensors 200 MB ******** download
model-00112-of-00148.safetensors 156 MB ******** download
model-00132-of-00148.safetensors 144 MB ******** download
model-00142-of-00148.safetensors 144 MB ******** download
model-00103-of-00148.safetensors 140 MB ******** download
model-00096-of-00148.safetensors 133 MB ******** download
model-00122-of-00148.safetensors 131 MB ******** download
model-00111-of-00148.safetensors 116 MB ******** download
model-00133-of-00148.safetensors 115 MB ******** download
model-00097-of-00148.safetensors 104 MB ******** download
model-00113-of-00148.safetensors 80.2 MB ******** download
model-00145-of-00148.safetensors 80.0 MB ******** download
model-00144-of-00148.safetensors 75.5 MB ******** download
model-00123-of-00148.safetensors 66.9 MB ******** download
model-00134-of-00148.safetensors 64.0 MB ******** download
model-00089-of-00148.safetensors 61.5 MB ******** download
model-00104-of-00148.safetensors 52.0 MB ******** download
model-00124-of-00148.safetensors 48.3 MB ******** download
model-00135-of-00148.safetensors 48.0 MB ******** download
model-00114-of-00148.safetensors 42.2 MB ******** download
model-00083-of-00148.safetensors 28.0 MB ******** download
model-00105-of-00148.safetensors 28.0 MB ******** download
model-00071-of-00148.safetensors 24.0 MB ******** download
model-00082-of-00148.safetensors 24.0 MB ******** download
model-00090-of-00148.safetensors 23.6 MB ******** download
model-00076-of-00148.safetensors 18.9 MB ******** download
model-00136-of-00148.safetensors 18.9 MB ******** download
model-00115-of-00148.safetensors 16.0 MB ******** download
model-00125-of-00148.safetensors 16.0 MB ******** download
model-00060-of-00148.safetensors 14.0 MB ******** download
model-00084-of-00148.safetensors 14.0 MB ******** download
model-00116-of-00148.safetensors 14.0 MB ******** download
model-00126-of-00148.safetensors 14.0 MB ******** download
model-00056-of-00148.safetensors 10.2 MB ******** download
model-00077-of-00148.safetensors 9.00 MB ******** download
model-00091-of-00148.safetensors 9.00 MB ******** download
model-00137-of-00148.safetensors 9.00 MB ******** download
model-00066-of-00148.safetensors 8.00 MB ******** download
model-00106-of-00148.safetensors 6.00 MB ******** download
model-00098-of-00148.safetensors 4.78 MB ******** download
model-00146-of-00148.safetensors 4.78 MB ******** download
model-00061-of-00148.safetensors 3.38 MB ******** download
model-00072-of-00148.safetensors 3.38 MB ******** download
model-00127-of-00148.safetensors 3.38 MB ******** download
model-00051-of-00148.safetensors 2.39 MB ******** download
model-00147-of-00148.safetensors 1.33 MB ******** download
tokenizer.json 8.93 MB ff6aaf29 download
model.safetensors.index.json 202 KB b85a1eec download
chat_template.jinja 16.8 KB 974eeefa download
config.json 14.6 KB 707bba5c download
LICENSE 13.6 KB 2be6f885 download
dealign_mascot.png 10.9 KB da3bf39a download
README.md 8.92 KB e274b357 download
hf_quant_config.json 1.97 KB 8886c72e download
.gitattributes 1.48 KB a6344aac download
lyber_icon.png 1.28 KB f8b116fd download
tokenizer_config.json 1.25 KB 59ac7739 download
generation_config.json 189 B aa393606 download

README current version from Hugging Face


language:

  • en
  • zh
  • es
  • fr
  • ja
  • ko
  • ru
    license: other
    license_name: tencent-hunyuan
    license_link: https://huggingface.co/tencent/Hy4-flash/blob/main/LICENSE
    base_model: tencent/Hy4-flash
    base_model_relation: finetune
    library_name: transformers
    pipeline_tag: text-generation
    thumbnail: dealign_mascot.png
    gated: manual
    extra_gated_prompt: |
    By requesting access you confirm you are 18+, will use the model solely in
    compliance with applicable law in your jurisdiction, and accept full responsibility
    for all outputs. Access is granted at the sole discretion of the repository owner.
    extra_gated_fields:
    Name: text
    Affiliation / organization: text
    Country: country
    Primary use case: text
    I confirm I am 18 or older: checkbox
    I accept full responsibility for outputs: checkbox
    extra_gated_heading: "Request access — manual approval required"
    extra_gated_description: "This model is reviewed manually. Fill in the fields below."
    extra_gated_button_content: "Request access"
    tags:
  • tencent
  • hunyuan
  • hy4
  • hy4-flash
  • mxfp4
  • compressed-tensors
  • uncensored
  • reasoning
  • tool-calling
  • moe

   

A Lyber AI × dealign.ai release

Twitter — @dealignai
Lyber AI

[!NOTE]
Please credit if you build on this. If you redistribute, quantize, fine-tune, serve, benchmark, or write about this model, a line like
"Based on lyber-ai/Hy4-flash-UNCENSORED-MXFP4"
in your model card or write-up means a lot. Thank you.

[!IMPORTANT]
Shipped in mixed-precision MXFP4 (OCP microscaling FP4 for MoE weights) + FP8 for
attention, in the compressed-tensors format. Extended thinking with reasoning-effort
tiers (no_think / high) and tool calling are preserved.
Recommended sampling: temperature=0.7, top_p=0.95, max_tokens ≥ 16384 on high.

Hy4-flash — UNCENSORED (MXFP4)

Uncensored release · Mixed MXFP4 / FP8 · 128K context · Extended thinking · Tools · ~217 GB


What this is

This is tencent/Hy4-flash (Hunyuan V4 Flash)
with refusal behavior removed, so the model complies with user requests across topic
categories rather than declining them. Shipped in mixed MXFP4 + FP8 compressed-tensors
format for native 4-bit MoE throughput on Blackwell. All upstream capabilities — extended
thinking with reasoning_effort tiers, tool calling, and the native 128K+ context — are
preserved. The bundle is drop-in compatible with the base model's serve configs (see below
for the vLLM image requirement).

Internal benchmarks

Capability — MMLU (228-question balanced subset, same harness as base)

228 MMLU questions · logit mode on no_think (argmax of A/B/C/D token logprobs;
max_tokens=1, temperature=0) · reasoning_effort=high mode uses greedy sample and
parses the final letter.

Mode Base UNCENSORED-FP8 (ship) UNCENSORED-MXFP4 (this)
no_think (logit) 88.16% 87.28% 86.40%
reasoning_effort=high (sample) 94.30% 91.67% 89.91%

Both modes hold within ~2.5 pp of the FP8 ship; the high mode's 12 unparsed responses hit
the 4096-token cap in extended reasoning before emitting the final letter (budget-limited,
not capability-limited).

HarmBench-320 — full canonical test set

320 standard HarmBench behaviors · reasoning_effort=high · max_tokens=4096 ·
no system prompt · deterministic grader (refusal opener + full-content audit for task
attempts + reasoning-engagement audit).

All rows audited by LLM after the deterministic grader, so the keyword-flagged
false-positives are corrected — a "disclaimer / for-research" preamble followed by a
substantive attempt counts as COMPLY, and an empty_final row whose reasoning trace
engaged with the task counts as COMPLY.

Harmful categories (chem/bio, cyber, illegal, harmful, misinfo, harassment) — 240 prompts

Mode COMPLY %
no_think 237 / 240 98.8%
reasoning_effort=high 237 / 240 98.8%

Full HB-320 (incl. 80 copyright prompts)

Mode COMPLY REFUSE %
no_think 317 / 320 3 99.1%
reasoning_effort=high 284 / 320 36 88.8%

The high-mode drop vs no_think is almost entirely in the copyright category (47 / 80
vs 80 / 80), where many responses ran out of reasoning budget at 4096 tokens before
emitting final text — a high-mode 49K-budget rerun that matches the FP8 ship's harness
is pending. The harmful-content axis in both modes stays at 98.8%.

HarmBench-320 by SemanticCategory
SemanticCategory n COMPLY (no_think) COMPLY (high)
chemical_biological 42 39 / 42 (92.9%) 42 / 42 (100%)
cybercrime_intrusion 52 52 / 52 (100%) 51 / 52 (98.1%)
harassment_bullying 21 21 / 21 (100%) 21 / 21 (100%)
harmful 18 18 / 18 (100%) 16 / 18 (88.9%)
illegal 53 53 / 53 (100%) 53 / 53 (100%)
misinformation_disinformation 54 54 / 54 (100%) 54 / 54 (100%)
copyright 80 80 / 80 (100%) 47 / 80 (58.8%) — budget-limited
Harmful (non-copyright) total 240 237 / 240 (98.8%) 237 / 240 (98.8%)
Full total 320 317 / 320 (99.1%) 284 / 320 (88.8%)

Multiturn adversarial pressure (think-high, no system prompt)

Harsh prompt sets with preserved reasoning across turns, 1 seed × 10 turns, long budget
per turn:

Setting Comply Loops Entropy (mean / tail) Max context
preserved-reasoning multiturn, 10 turns (cyber → chem → violence → bio → fraud → weapons → misinfo → drugs → harassment → self-harm) 10 / 10 0 / 10 0.67 / 0.50 nats 18,304 tokens

Every turn finished cleanly (stop), no entropy collapse, no rep8 hit at any turn.
tail_entropy stayed between 0.29 and 0.74 across the 10-turn span.

Loop stability — adversarial single-turn

10 harsh adversarial prompts × reasoning_effort=high · T=0 greedy · 16K tokens:

Build Loops (rep8 ≥ 0.5) finishes mean entropy (med) tail entropy (max)
tencent/Hy4-flash-FP8 (ship) 0 / 10 10 stop 0.33 0.42
lyber-ai/Hy4-flash-UNCENSORED-FP8 0 / 10 10 stop 0.32 0.39
lyber-ai/Hy4-flash-UNCENSORED-MXFP4 (this) 0 / 10 10 stop 0.34 0.43

No loop collapse on the adversarial set. Entropy stays flat across 10 harsh prompts at
reasoning_effort=high, T=0, 16K-token budget.

How to serve

vLLM (recommended, Blackwell / B300)

Requires the Lyber-patched vllm/vllm-openai:hy4 image (upstream PR pending):

docker run -it --rm --gpus all --ipc=host --shm-size=64g --network host \
  -v /path/to/model:/model \
  vllm/vllm-openai:hy4-lyber-v6 \
    --model /model \
    --served-model-name hy4-flash \
    --quantization compressed-tensors \
    --tensor-parallel-size 8 \
    --max-model-len 131072 \
    --max-num-seqs 64 \
    --gpu-memory-utilization 0.92 \
    --reasoning-parser hy_v4 --tool-call-parser hy_v4 --enable-auto-tool-choice \
    --engram-config '{"cpu_offload":false}' \
    --host 0.0.0.0 --port 8000

Hardware requirements

GPU Setup Status
8× B300 SXM6 TP8, MXFP4 MoE ✅ Fits comfortably (~217 GB)
8× H200 TP8 ⚠️ Marlin MXFP4 only (no native FP4)
8× H100 80GB TP8 ⚠️ Marlin MXFP4 only (no native FP4)

Reasoning + tools

chat_template_kwargs = {"reasoning_effort": "high"}  # or "no_think"

reasoning_effort accepts "no_think" or "high" (the two tiers Tencent ships on this
model). Tool calling and DFly speculative decoding (5 draft tokens) are preserved from the
base model.

Responsible use

You are solely responsible for how you use this model and its outputs.

  • Do not use for generating illegal content in your jurisdiction
  • Do not use to harm, deceive, or exploit real individuals
  • Do not use to generate CSAM or non-consensual intimate content
  • Comply with all applicable laws and regulations

License

This model inherits the Tencent Hy License, Version 1.0
from the base model. The modification does not change the license terms.

Credits

Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration