DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2)

Overview

A community MoEspresso V2 requant of DeepSeek's code-focused V4 Flash — MoE with 284B total/13B active params, 1M context, MIT-licensed, ~165GB VRAM at Q4.

When to pick this model

• Code-focused agentic workloads that want DeepSeek V4 Flash Coder without the full official footprint
• Multi-GPU setups that can't fit the full 176GB official build but can handle ~165GB
• Teams comfortable with community requants rather than official DeepSeek releases
• Long-context coding tasks needing up to 1M tokens

VRAM requirements by quantization

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

In practice, DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) is server-class even at Q4_K_M (165 GB). Stepping up to Q8_0 nearly doubles the footprint to 304 GB, and unquantized FP16 weights take 568 GB — plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity.

Without a GPU, DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) needs roughly 369 GB of system RAM to run on CPU via llama.cpp or Ollama — workable for background jobs, but far slower than GPU inference. Throughput estimates from our compatibility engine: around 18 tokens/sec on entry-level GPUs, on the order of 28 tokens/sec on a mid-range card, and up to 45 tokens/sec on high-end hardware — assuming the chosen quantization fully fits in VRAM.

What hardware do you need

The table below matches DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) to common GPU memory tiers, using the highest-fidelity quantization that fully fits each card class. Spilling layers to system RAM works but costs most of the speed, so size your card to the quantization you actually want to run.

Which GPU should you buy to run DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2)?

To run DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) locally at Q4, you need ~165 GB of VRAM. The best value for this is a Apple Mac Studio (64+ GB unified memory).

Strengths

• Native 1M-token context
• Efficient MoE: only ~13B of 284B parameters active, giving decent throughput for the size
• Purpose-built for code and coding agents
• Permissive MIT license

Limitations

• Still ~165GB VRAM at Q4 — needs multiple GPUs or a large RAM pool
• A community repack, not an official DeepSeek build
• No official Ollama tag — HuggingFace install only

Typical workloads

In our catalog grid, DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) is filed under Code, Coding Agents, Long Context 1M — the use cases where its size/quality trade-off makes the most sense. Its tags translate to concrete workloads: code generation and review (pair it with an editor integration like Continue.dev or Cline).

The 1024k-token context window is large enough to hold entire codebases' worth of files or long reports in a single prompt, which is what makes local RAG and document analysis practical. The MIT license is permissive, so shipping it inside a commercial product raises no special legal questions.

Architecture & training

Architecture: Sparse MoE · 284B total parameters, ~13B active per token · 1,048,576-token context window (1M) · quantized "MoEspresso V2" repack (~56.8GB)

Training: Community requant/repack (steadfastgaze) of the code-focused DeepSeek V4 Flash 0731. Official DeepSeek base; MoEspresso V2 format optimized for memory footprint. Training details not published.

Quick start

# HuggingFace : steadfastgaze/DeepSeek-V4-Flash-0731-Coder-56.8GB-MoEspressoV2

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.

Similar models worth comparing

Frequently asked questions

How much VRAM does DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) need?

At the recommended Q4_K_M quantization, DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) needs about 165 GB of VRAM. Q8_0 takes 304 GB, and unquantized FP16 weights take 568 GB.

Can DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) run without a GPU?

Yes — with roughly 369 GB of system RAM it runs CPU-only through llama.cpp or Ollama. Expect a fraction of GPU speed, which is fine for background or batch jobs but slow for interactive chat.

What context window does DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) support?

DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) supports a 1024k-token context window (1,048,576 tokens).

Can I use DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) commercially?

Yes. DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) is released under MIT, a permissive open-source license that allows commercial use, modification and redistribution.

How fast is DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) on consumer hardware?

Our compatibility engine estimates on the order of 28 tokens/sec on a mid-range GPU and up to 45 tokens/sec on high-end cards, assuming the quantization fully fits in VRAM.

Which quantization of DeepSeek V4 Flash Coder 284B-A13B (MoEspresso V2) should I download first?

Start with Q4_K_M (165 GB) — the standard size/quality sweet spot. Step up to Q5_K_M or Q8_0 only if you have VRAM headroom. It does not fit a single 24 GB consumer card — plan for multi-GPU or server hardware.