[Desktop] weights failed to load: Qwen3.8-Flash-Next NVFP4

[Ramachandrajoshi]

Before you start

• I have checked the supported models and this checkpoint is not listed there. A listed checkpoint that fails is a bug; use a Bug report instead.
• I have read the Roadmap and this model is not already planned there.
• I have read the FAQ.
• I have searched existing issues and found no duplicate.
• This is not a GGUF checkpoint. GGUF support is on the Roadmap; please do not open an issue for it.

Hugging Face link

https://huggingface.co/RadixArk/Qwen3.8-Flash-Next-NVFP4

Is the model architecture already supported

Yes, but this checkpoint or quantization does not load

Is the quantization already supported

Yes, but this checkpoint's weight format does not load

What happens when you load it

>>>>>>>> PASTE THE FULL SERVER LOG HERE <<<<<<<<
(copied with the "Copy server log" button in the failure dialog, or Logs -> Server status -> Copy)

WeightLoadError: OSError: [WinError 1455] The paging file is too small for this operation to complete

Anything else

Desktop 0.2.0-beta.21; engine 0.1.3+gb332d3a97; NVIDIA GeForce RTX 5060 Ti, 15.9 GiB, driver 610.88; AMD Ryzen 7 5800X 8-Core Processor, 95.9 GiB

[benhadad]

OP issue does not seem to be with the program but local computer setup.

OSError: [WinError 1455] The paging file is too small for this operation to complete is an explicit Windows operating system error, not an internal bug in FreeToken or the model's architecture.

What causes it
Committed Virtual Memory Limit: When Windows tries to allocate virtual address space or memory-map huge model weight files into RAM/VRAM, the total committed memory (Physical RAM + pagefile.sys) exceeds the allowable commit limit.

Disk Space Constraints: Even though the user has ~96 GB of RAM, Windows cannot dynamically expand the paging file if the drive hosting pagefile.sys (usually C:) is nearly full.

Fixed Paging File Size: If the user previously set a manual, fixed size for the Windows paging file (or disabled it entirely), large memory-mapped allocations during weight loading will immediately fail with error 1455.

Recommended fix for the user
Free up drive space: Ensure the system drive (and whichever drive FreeToken caches models to) has ample headroom (at least 50–100 GB free).

Increase or automate pagefile size:

Open sysdm.cpl (System Properties) -> Advanced tab.

Under Performance, click Settings -> Advanced tab -> Virtual memory -> Change...

Check Automatically manage paging file size for all drives, or manually assign a custom size with a larger maximum (e.g., 32 GB to 64 GB) on an SSD with sufficient free space.

Reboot Windows to apply the changes.

[trcwebdesign]

You could try my fork
https://github.com/trcwebdesign/FreeToken
Windows NVFP4 Fork Enhancements
This fork adds Windows-focused support for serving large Qwen3.8 and Gemma4 NVFP4/GGUF checkpoints on consumer hardware.

Disk-tier MoE offload
--moe-disk-tier on --expert-ram-experts 128 --disk-fetch-workers 16

[jason-fxz]

@Ramachandrajoshi Please provide FULL SERVER LOG: Logs -> Server status -> Copy