license: apache-2.0
base_model: DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored
pipeline_tag: text-generation
library_name: transformers
description: "NVFP4 safetensors checkpoint of DavidAU's Ultimate-Heretic Turbo-Brilliance-50x-AGI-NEO tune of Qwen3.8-27B: 256 packed NVFP4 projections (MLP plus full-attention QKVO) with the DeltaNet path, MTP head, vision tower, and lm_head in BF16. MTP in every name. Blackwell sm_120."
tags:
- nvfp4
- qwen3.8
- blackwell
- mtp
- speculative-decoding
- ultimate-heretic
- compressed-tensors
- llama.cpp
language: - en
- multilingual
quantization_config:
quant_method: compressed-tensors
format: nvfp4-pack-quantized
Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NVFP4
NVFP4 checkpoint of DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored, DavidAU's Ultimate-Heretic Turbo-Brilliance-50x-AGI-NEO tune of Qwen3.8-27B, quantized from the BF16 release with a conservative 256-tensor recipe.
HF repo: esatapedico/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NVFP4
Companion serving build: the 400-tensor W4A16 checkpoint lives in esatapedico/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NVFP4-W4A16-T400 with its own card.
NVFP4 here means W4A16 with FP8 E4M3 scales, group size 16, weight only, datafree (no calibration corpus). The quantized set covers the MLP on all 64 layers plus Q, K, V, O on the 16 full-attention layers (256 packed NVFP4 tensors, verified intact at save time). The DeltaNet path stays BF16 in this checkpoint and is normalized to NVFP4 in the GGUF tier builds (448-tensor backbone). Vision tower, MTP head, lm_head, and token embeddings stay at full precision.
Follow along & support
I post updates on new conversions, benchmarks, and what I'm working on over on my site. Follow along there to keep up with new releases and the work in progress. If you'd like to support more of it, you will find the options there. I do this on consumer hardware and like seeing how far it goes. More is on the way.
🌐 esatapedico.com. Updates, work-in-progress, and ways to support.
At a glance
| Field | Value |
|---|---|
| Format | compressed-tensors nvfp4-pack-quantized |
| Quantization | W4A16 ("NVFP4A16"), group size 16, FP8 E4M3 scales, weight only |
| Calibration | none (datafree) |
| Quantized | MLP on all 64 layers + Q/K/V/O on the 16 full-attention layers (256 packed NVFP4 tensors) |
| Kept in BF16 | DeltaNet path, vision tower, MTP head (15 tensors), lm_head, token embeddings, norms, scales |
| Base | DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored (27B dense hybrid, Apache-2.0) |
Checkpoint
Two weight shards plus config and tokenizer files:
| File | Size | SHA-256 | Notes |
|---|---|---|---|
model-00001-of-00002.safetensors |
19,967,491,332 bytes (19.97 GB) | 37579ec70627c7a7e061640acca5c9f2c739ae46f55bc12015c811b27c14b32c |
weight shard 1 of 2, packed NVFP4 plus BF16 remainder |
model-00002-of-00002.safetensors |
8,584,268,068 bytes (8.58 GB) | 0055b7f48af07a7543ff6405ab35042cec6b772e7e492384e1550db0e7d50cf3 |
weight shard 2 of 2, packed NVFP4 plus BF16 remainder |
model.safetensors.index.json |
167,561 bytes | 3e7d9cbce9925dfea9409e1114c3f1e96ab892fa5f322421483e83574cd9e39e |
weight map, total size 28,551,759,408 |
config.json |
24,468 bytes | b586120f3c8bb23517954a35e6e41458d4c21453959756e612257aedc190d46c |
architecture plus quantization_config |
tokenizer.json |
19,989,506 bytes (20.0 MB) | 2f4dd754486e96054d0c841ddde816748be0116cbbc8ba8b5311ae4853432a21 |
tokenizer |
tokenizer_config.json |
1,569 bytes | 13e78188447d1a36c7bc332a2c20f6eadb9823c255c5565c8107cf3e055111ca |
tokenizer config |
chat_template.jinja |
248,063 bytes | 73c64a8ea9af4491bda8b4f071964cf55dca998e8dbecea1538f1a0b8341dcf0 |
chat template |
generation_config.json |
219 bytes | 2c946d7c1ecbdfcb279a879795b7c28323119ade9b90b19f3170e0e1da2da8fa |
generation defaults |
recipe.yaml |
251 bytes | 557d35364be80168874c123db83ec1dcba4766daf94c8706cdcd24a59e6e47ea |
quantization recipe used |
Measurements
- Packed census: 256 of 256
Lineartargets quantized and packed, verified intact at save time. - MTP check: all 15 MTP tensors present in BF16, bit-identical to the source (14 versus the source shard plus
mtp.fc.weightversus the source MTP restoration file). - Preflight check: PASS.
- Total weight size: 28,551,759,400 bytes (28.55 GB) in two shards.
Usage
This checkpoint is the quantization source for GGUF builds. The serving path is a converted GGUF file:
convert_hf_to_gguf.py --outtype auto
which keeps the native NVFP4 tensors intact. Requires a recent llama.cpp with NVFP4 (GGML type 40) support.
How this was made
- Quantized with a datafree NVFP4A16 recipe: targets
Linear, ignore listlm_head, visual tower, fulllinear_attnpath. Sequential run, CPU only, no calibration corpus. - Packed 256 NVFP4 tensors. The 15 MTP tensors were restored to BF16 after quantization and verified bit-identical.
- Saved as two safetensors shards with config, tokenizer, and recipe files.
Attribution & provenance
This is a derivative work built entirely from existing Apache-2.0 artifacts. Nothing here was trained or fine-tuned. This work covers the numeric conversion only and makes no ownership claims over the model. Credit belongs to:
- Alibaba / Qwen team for the base model family (Apache-2.0): the 27B dense hybrid architecture this tune builds on.
- DavidAU for the tune itself, Qwen3.8-27B-Ultimate-Heretic-Uncensored (Apache-2.0), including the Turbo-Brilliance-50x-AGI-NEO system this checkpoint is named for.
- This repo's author for the NVFP4 conversion and verification only.
License
Apache-2.0, identical to every upstream artifact. The base model license governs. Quantization is a transformation, not new training. When redistributing, please retain attribution to Qwen (Alibaba) and DavidAU as above.
"Qwen" is a trademark of Alibaba. Trademarks are used here only to identify upstream models. This repository is not affiliated with, sponsored by, or endorsed by Alibaba or DavidAU.
Card written with AI assistance.