---
title: Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF
canonical_url: "https://www.modelscope.ai/models/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF"
md_url: "https://www.modelscope.ai/models/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF.md"
repository: satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF
last_updated: 2026-07-06
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
base_model:
  - HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive
base_model_relation: quantized
library_name:
  - gguf
downloads: 748
stars: 2
tags:
  - gguf
  - long-context
  - yarn
  - qwen3.6
  - uncensored
  - mtp
  - speculative-decoding
  - vision
  - llama.cpp
  - ollama
---

# Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF

> Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF - An open-source model by satgeze on ModelScope. Qwen3.6-35B Uncensored: 1M Context + MTP + Vision, one file

satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF is a text-generation model on ModelScope. licensed under apache-2.0. derived from HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive.

- **Repository**: satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF
- **License**: apache-2.0
- **Tasks**: text-generation
- **Base model**: HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive
- **Tags**: gguf, long-context, yarn, qwen3.6, uncensored, mtp, speculative-decoding, vision, llama.cpp, ollama
- **Downloads**: 748
- **Stars**: 2
- **Last updated**: 2026-07-06

Source: https://www.modelscope.ai/models/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF

---

<img src="banner.jpeg" width="720"/>

# Qwen3.6-35B Uncensored: 1M Context + MTP + Vision, one file

[HauhauCS's Aggressive uncensored build](https://huggingface.co/HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive) of Qwen3.6-35B-A3B, extended to a **1,048,576-token context** and fused with the official MTP speculative-decoding layer. Every claim on the banner is measured, not inherited. Verification data ships in this repo.

<table>
<tr>
<th style="background:#e8590c;color:#fff;padding:8px 14px;">Capability</th>
<th style="background:#e8590c;color:#fff;padding:8px 14px;">Status</th>
<th style="background:#e8590c;color:#fff;padding:8px 14px;">Evidence</th>
</tr>
<tr><td><b>1M context</b></td><td>Certified July 6, 2026</td><td><b>70/70 needles</b>, full ladder 64K to 1M, 10 depths per rung, f16 KV, temp 0</td></tr>
<tr><td><b>MTP speculative decoding</b></td><td>Grafted and vetted</td><td>284.6 to <b>380.1 tok/s (+33.6%)</b>, draft acceptance <b>0.864</b>, RTX 5090</td></tr>
<tr><td><b>Vision</b></td><td>Verified</td><td>mmproj tower reads image text and identifies objects correctly</td></tr>
<tr><td><b>Uncensored</b></td><td>HauhauCS Aggressive</td><td>Trunk weights bit-identical to the HauhauCS release</td></tr>
</table>

## Needle-in-a-haystack certification

<img src="qwen36_heatmap.png" width="640"/>

Native context is 262,144. YaRN rope scaling (factor 4) extends to 1,048,576, and retrieval stays perfect across the entire extension: every rung from 64K to 1M scored 10/10 at depths from 5 to 95 percent. Raw per-needle records are in `results.jsonl`.

## MTP: how it got here and what it does

<img src="qwen36_mtp_speedup.png" width="560"/>

Qwen3.6 ships a multi-token-prediction layer in the official checkpoint, but uncensored community trunks usually drop it. This build grafts the official MTP layer (via [Unsloth's MTP GGUF](https://huggingface.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF)) back onto the uncensored trunk at the GGUF tensor level. The draft head then predicts ahead and the trunk verifies every token, so **output is identical to standard decoding, only faster**.

Measured on the uncensored trunk (RTX 5090, 16K ctx): 284.6 tok/s standard, 380.1 tok/s with `--spec-type draft-mtp` (+33.6 percent). Draft acceptance 0.864 with mean accepted run 3.59, which is higher than we measured on the official trunk (0.816). A 262K needle test with speculation active scored 10/10, confirming the graft changes nothing about retrieval.

## Files

| File | Size | What it is |
|---|---|---|
| `qwen3.6-35b-uncensored-1M-MTP-Q4_K_M.gguf` | 21.7 GB | The one-file package: uncensored trunk + 1M rope + MTP layer |
| `mmproj-qwen36-hauhau-f16.gguf` | 899 MB | Vision tower, attach at runtime |
| `qwen36_heatmap.png`, `qwen36_mtp_speedup.png`, `results.jsonl` | small | Verification evidence |

The plain non-MTP 1M trunk and other quants live on the ModelScope mirror.

## Every file, every mirror

Nothing was discontinued: every quant is one click away. Hugging Face carries the curated picks, ModelScope always carries everything, and Ollama serves ready-to-run tags.

On Ollama every tag ships with the vision tower bundled and the 1M rope metadata baked in.

| File | Size | Hugging Face | ModelScope | Ollama |
|---|---|---|---|---|
| `mmproj-qwen36-hauhau-f16.gguf` | 899 MB | [download](https://huggingface.co/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF/resolve/main/mmproj-qwen36-hauhau-f16.gguf) | [download](https://www.modelscope.ai/models/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF/resolve/master/mmproj-qwen36-hauhau-f16.gguf) | bundled in every tag |
| `qwen3.6-35b-uncensored-1M-MTP-Q4_K_M.gguf` | 21.7 GB | [download](https://huggingface.co/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF/resolve/main/qwen3.6-35b-uncensored-1M-MTP-Q4_K_M.gguf) | [download](https://www.modelscope.ai/models/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF/resolve/master/qwen3.6-35b-uncensored-1M-MTP-Q4_K_M.gguf) | [`ollama run satgeze/qwen36-35b-uncensored-1m`](https://ollama.com/satgeze/qwen36-35b-uncensored-1m) |
| `qwen3.6-35b-uncensored-1M-Q4_K_M.gguf` | 21.2 GB | on ModelScope | [download](https://www.modelscope.ai/models/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF/resolve/master/qwen3.6-35b-uncensored-1M-Q4_K_M.gguf) | [`ollama run satgeze/qwen36-35b-uncensored-1m:q4_k_m-no-mtp`](https://ollama.com/satgeze/qwen36-35b-uncensored-1m) |

## Run with Claude Code (or any Anthropic-API agent)

llama-server natively speaks the Anthropic messages protocol, so Claude Code can use this model directly. Two integration fixes are required, both included here:

```bash
llama-server -m qwen3.6-35b-uncensored-1M-MTP-Q4_K_M.gguf \
  -c 262144 -np 2 --cache-reuse 256 -ngl 99 --jinja \
  --chat-template-file chat_template_claude_code.jinja \
  --spec-type draft-mtp --spec-draft-n-max 3 \
  --mmproj mmproj-qwen36-hauhau-f16.gguf \
  --port 8040 --host 0.0.0.0
```

Then in the client shell:

```bash
export ANTHROPIC_BASE_URL=http://<server-ip>:8040
export ANTHROPIC_AUTH_TOKEN=local
export ANTHROPIC_MODEL=qwen36-35b-uncensored-1m-mtp
claude
```

Why the flags: `chat_template_claude_code.jinja` is the embedded template with one strictness check relaxed (the original raises an exception on the multiple system blocks agent harnesses send). `-np 2` matters because Claude Code fires background side-requests; with a single slot they evict your main generation. Measured on an RTX 5090: 330 to 380 tok/s decode with speculation active, tool calls and vision working.

## How this was built

1. **1M context**: YaRN rope-scaling metadata (factor 4.0, original context 262,144) written directly into the GGUF header with gguf-py. No weights change, no fine-tuning, no runtime flags needed: llama.cpp and Ollama read the baked metadata.
2. **MTP graft**: the official MTP layer tensors (block 41 of the donor GGUF) appended onto the uncensored trunk, `block_count` and `nextn_predict_layers` metadata updated to match. GGUF surgery, not training.
3. **Certification**: multi-needle retrieval harness (10 needles per rung at depths 5 to 95 percent, temperature 0, seeded to bust prompt caches) run against llama-server with f16 KV cache. Certification is only ever done on f16 KV; quantized KV is a labeled budget option, never the baseline.

## Run it

llama.cpp (full speed, MTP active):

```bash
llama-server -m qwen3.6-35b-uncensored-1M-MTP-Q4_K_M.gguf \
  -c 1048576 -np 1 --jinja \
  --spec-type draft-mtp --spec-draft-n-max 3 \
  --mmproj mmproj-qwen36-hauhau-f16.gguf
```

Ollama (1M and vision work; MTP tensors load harmlessly but Ollama has no speculative decoding yet, so no speed gain there):

```
FROM ./qwen3.6-35b-uncensored-1M-MTP-Q4_K_M.gguf
RENDERER qwen3.5
PARSER qwen3.5
PARAMETER num_ctx 262144
```

The RENDERER and PARSER lines matter: imported GGUFs without a native renderer hit template bugs under tool-heavy use (for example as a coding-agent backend). Raise `num_ctx` as memory allows. Measured footprints: 44 GB total at 1M f16 KV on a 128 GB Mac (100 percent GPU, 72 tok/s decode); a 32 GB card holds roughly 262K fully resident.

## Honest notes

- Uncensoring quality versus the official model has not been independently benchmarked here. For capability benchmarks of the base model, see the [official Qwen3.6-35B-A3B card](https://huggingface.co/Qwen/Qwen3.6-35B-A3B); weights here are bit-identical to the HauhauCS release apart from rope metadata and the added MTP layer, so base capability carries over modulo Q4_K_M quantization.
- The MTP layer came from the official checkpoint and pairs with this trunk architecture; acceptance 0.864 was measured on real coding prompts, and speculative decoding never changes outputs by construction.

## Credits

Base model: [Qwen](https://huggingface.co/Qwen/Qwen3.6-35B-A3B) (Apache-2.0), including the official MTP layer. Uncensoring and trunk quant: [HauhauCS](https://huggingface.co/HauhauCS). MTP GGUF packaging: [Unsloth](https://huggingface.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF). 1M YaRN extension, MTP graft, and certification: [SatGeze](https://huggingface.co/satgeze).

Mirrors: [Hugging Face](https://huggingface.co/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF) | [ModelScope](https://www.modelscope.ai/models/satgeze/Qwen3.6-35B-Uncensored-HauhauCS-1M-GGUF). Tooling: [github.com/satindergrewal/aviary-1m](https://github.com/satindergrewal/aviary-1m).