Qwen3.6‑35B‑A3B‑Uncensored‑HauhauCS‑Aggressive GGUF 加载报错 qwen35moe architecture not supported

[yunxilyf]

Before you start

• I have read the FAQ and my problem is not answered there.
• I have read the Roadmap and this is not already planned there.
• I have searched existing issues and found no duplicate.
• I have restarted the Desktop app to pick up the latest update and the problem still happens.

What happened

Observation:

• The NVFP4 variant works fine, proving FreeToken underlying llama.cpp does support qwen35moe architecture.
• The broken local‑imported Q4_K_M model shows ctx: 0 in model list, while the working NVFP4 shows correct ctx: 262144.
• Both files are Qwen3.6‑35B‑A3B MoE (qwen35moe).

To Reproduce

1. Local‑import GGUF: Qwen3.6‑35B‑A3B‑Uncensored‑HauhauCS‑Aggressive.Q4_K_M.gguf
2. Click start button to launch model
3. Engine popup: 引擎故障 ValueError: GGUF architecture 'qwen35moe' is not supported

Expected behavior

Should load the qwen35moe MoE model same as the official NVFP4 quantized version.

Screenshots

(paste your two screenshots here)

Environment

• OS: Windows
• FreeToken version: [fill your version]
• Working model: Qwen3.6‑35B‑A3B NVFP4 (19.5GiB, MoE tag shown, ctx=262144)
• Broken model: Qwen3.6‑35B‑A3B‑Uncensored‑HauhauCS‑Aggressive Q4_K_M (19.7GiB, local‑imported, ctx=0)

Additional context

Hypothesis:

1. The secondary‑tuned GGUF file may have corrupted GGUF header metadata, even though general.architecture="qwen35moe".
2. FreeToken / backend parser somehow rejects this specific file, even though same‑architecture NVFP4 works.
3. The ctx:0 status indicates metadata fields are not correctly parsed from this local imported GGUF.

Maybe FreeToken’s GGUF metadata validator has strict checks that reject this specific file, while upstream llama.cpp can handle it.

Desktop app version

0.2.0 beta 17

OS

Windows 11

OS details

No response

GPU and driver

4070ti surper

CPU and system RAM

9950x 64G

Checkpoint

Qwen3.6‑35B‑A3B‑Uncensored‑HauhauCS‑Aggressive Q4_K_M

Model settings

No response

Engine log

Anything else

No response