Fx-Work-1028-MXFP4-MoE-GGUF

Fx-Work-1028 is an improved multimodal MoE model built on Qwen3.6-35B-A3B, with approximately 3B active parameters, designed for professional workflows involving documents, spreadsheets, business data, and multi-file tool use. It retains a 262K context window and targets realistic, long-horizon work environments. Across five benchmarks, it performs comparably to GLM-5.2, scoring 34.34 on JobBench Main, 43.33 on AutomationBench, and 70.1 on Claw-Eval pass³, while reaching 1328 Elo on GDPvalAA-v2.1 and 28.2 on APEX-Agents-AA. It improves over the original Fx-Work across all five metrics and leads models at or below 35B on four, with Occamy-1.0 slightly ahead on Claw-Eval. Supports SGLang, vLLM, and Transformers, with tool execution handled by the surrounding agent harness. Released under Apache-2.0, subject to the base model terms.

MTP — Speculative Decoding: Yes

Multi-Token Prediction (MTP) GGUF is a specialized GGUF model file format extension that integrates speculative decoding directly into the model weights to significantly accelerate local inference. Unlike traditional speculative decoding which requires a separate, smaller "draft" model, MTP GGUF files include additional output heads within the main model architecture that predict multiple future tokens in a single forward pass.

Model Files

File Name	Quant Type	File Size	File Link	Description
Fx-Work-1028.BF16.gguf	BF16	71.1 GB	Link	Full BF16 weights. Highest quality, largest file size.
Fx-Work-1028.MXFP4_MOE.gguf	MXFP4_MOE	20.3 GB	Link	Mixed precision / MoE optimized quantization. Highly efficient size for massive models, recommended if your inference engine supports it.
Fx-Work-1028.Q3_K_L.gguf	Q3_K_L	18.6 GB	Link	Lower quality but usable, good for low RAM availability.
Fx-Work-1028.Q5_K_M.gguf	Q5_K_M	25.3 GB	Link	High quality, recommended.
Fx-Work-1028.Q6_K.gguf	Q6_K	29.2 GB	Link	Very high quality, near perfect, recommended.
Fx-Work-1028.mmproj-bf16.gguf	mmproj-bf16	903 MB	Link	Multimodal projection file in BF16 format. Used for vision/language models.

MXFP4_MOE Test Run

Tested Fx-Work-1028.MXFP4_MOE.gguf on an NVIDIA RTX PRO 6000 Blackwell Server Edition GPU with 96 GB of VRAM, using llama-cli with full GPU offloading and MTP speculative decoding.

Performance Metrics

Metric	Value
Average Prompt Processing Speed	216.25 tokens/s
Average Generation Speed	297.38 tokens/s
Total Wall Time	56.4 seconds
Context Length	16,384 tokens
KV Cache (K)	q8_0
GPU	NVIDIA RTX PRO 6000 Blackwell Server Edition
GPU VRAM Capacity	96 GB
Peak GPU VRAM Usage	19.87 GiB
Peak GPU Utilization	81%
GPU Layers Offloaded	99 (-ngl 99)
Random Seed	42
Speculative Decoding	MTP (draft_n_max=2)
Test Runs	5

Test Shots

Shot	Test
Shot 1	C: Insert Node at End of Singly Linked List
Shot 2	Jogger and Train Passing Time
Shot 3	Two Pipes Filling a Tank
Shot 4	Students in Examination Rooms
Shot 5	Father and Son Ages

Run Inference

./build/bin/llama-cli \
  -m /path/to/models/Fx-Work-1028.MXFP4_MOE.gguf \
  --spec-type draft-mtp \
  --spec-draft-n-max 2 \
  -ngl 99 \
  -c 16384 \
  --cache-type-k q8_0 \
  --seed 42 \
  -n 8192 \
  -st \
  -p "Your prompt here"

Performance results reflect this specific hardware and inference configuration. Actual throughput may vary depending on the prompt, model, software build, and runtime settings.

llama.cpp

LLM inference in C/C++ — https://github.com/ggml-org/llama.cpp

---

Snapshots (5)

---

Quantized using llama.cpp by prithivMLmods