DeepSeek V4 Flash (MXFP4)
antirez's MXFP4 community GGUF of DeepSeek V4 Flash 0731 — the July 2026 re-post-trained 284B-parameter Mixture-of-Experts (~13B active per token) with 1M-token context and Compressed Sparse Attention. Routed experts use the MXFP4 micro-scaled 4-bit block format while attention projections, shared experts, router, embeddings, and output head stay at Q8_0 / F16, landing at ~156 GB — close to the native FP4 build's footprint and the highest-fidelity single-file ds4 variant. Needs a workstation or Mac Studio-class machine with ~200 GB of usable memory. Produced for antirez's `ds4` inference engine.
At a glance
- Parameters
- 284B
- Quantization
- MXFP4
- Context window
- 1000K tokens
- Approx. size
- 156.0 GB
- Engines
- llama.cpp
- License
- MIT (open)
- Version
- v1.0.0 , released 2026-08-01
- Family
- deepseek
Tuning
Defaults a gezel client applies out of the box, from real evaluation runs against this exact quantization.
- sampling.maxTokens
16384- reasoning.thinkingBudget
4096- reasoning.enableThinking
true- profiles
- thinking-general, thinking-coding, thinking-precise, instruct, creative
Model behaviors
Client-side behavior modules the catalog enables for this model (reasoning-tag handling, fabrication detection, prompt shaping).
- reasoning.strip-think-tags
- prompt.private-reasoning-guidance
- prompt.tool-cookbook-condensed
- prompt.meester-build-prelude
- mcp.compact-tool-schemas
- fabrication.detect-past-tense-no-tools
- turn.preamble-folding
Sources
- llama.cpp (GGUF)
- antirez/deepseek-v4-gguf ·
DeepSeek-V4-Flash-MXFP4Experts-F16HC-F16Compressor-F16Indexer-Q8Attn-Q8Shared-Q8Out-chat-v2-mxfp4-0731.gguf· MXFP4 - Upstream
- https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash
- Manifest
- View on GitHub
Tags
- deepseek
- reasoning
- agentic
- moe
- tools
- long-context
- large
- mxfp4
- fp4