DeepSeek V4 Flash (FP4)
DeepSeek V4 Flash — 284B-parameter Mixture-of-Experts (~13B active per token) with 1M-token context, Compressed Sparse Attention, and configurable Non-think / Think-High / Think-Max reasoning modes. On the `ds4` engine this is the July 2026 0731 re-post-trained checkpoint, which is substantially stronger at agentic and coding work than the original release. The llama.cpp (native FP4+FP8) and MLX sources still carry the original V4 Flash weights — upstream has not republished those builds for 0731 — so pick the ds4 source to get the newer checkpoint. Frontier-tier on-device reasoning for workstations and Mac Studio-class hardware — needs ~200 GB of usable memory.
At a glance
- Parameters
- 284B
- Quantization
- FP4-FP8
- Context window
- 1000K tokens
- Approx. size
- 164.6 GB
- Engines
- llama.cpp, MLX
- License
- MIT (open)
- Version
- v1.1.0 , 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)
- nsparks/DeepSeek-V4-Flash-FP4-FP8-GGUF ·
DeepSeek-V4-Flash-FP4-FP8-native.gguf· FP4-FP8 - MLX
- mlx-community/deepseek-ai-DeepSeek-V4-Flash-4bit
- Upstream
- https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash
- Manifest
- View on GitHub
Tags
- deepseek
- reasoning
- agentic
- moe
- tools
- fp4
- long-context
- large