DeepSeek DeepSeek Released 2024-05 codeMoEcompletion

DeepSeek Coder V2

DeepSeek Coder V2 open-source MoE code model. 16B active parameters with 236B total. Comparable to GPT-4-Turbo on code-specific tasks.

Best for strong 16B code model with MoE efficiency
Sizes 16B · 236B
Context 128K
License DeepSeek License
Min VRAM (default size, Q4) 10 GB
Rec VRAM 24 GB

What is DeepSeek Coder V2?

DeepSeek Coder V2 is DeepSeek's MoE open-weights language model. Released in 2024-05, it ships in 2 sizes (16B, 236B) and is licensed DeepSeek License. The most popular use case is strong 16B code model with MoE efficiency.

VRAM and hardware

The smallest 16B size needs at least 10 GB VRAM; the largest 236B needs around 10 GB.

Size Min VRAM (Q4_K_M) Recommended VRAM Tokens/sec on 3090
16B 10 GB 15 GB ~21 tok/s
236B 140 GB 210 GB ~11 tok/s

How to run DeepSeek Coder V2 locally

Option 1: Ollama (simplest)

ollama pull deepseek-coder-v2
ollama run deepseek-coder-v2

Option 2: Mullama (production)

mullama pull deepseek-coder-v2
mullama run deepseek-coder-v2

Option 3: llama.cpp (CLI)

# Download a GGUF from Hugging Face (search "deepseek-coder-v2 gguf")
./llama-cli -m deepseek-coder-v2.Q4_K_M.gguf -p "Hello, AI!"

Option 4: Python with Mullama or llama-cpp-python

from mullama import Model, Context
model = Model.load("deepseek-coder-v2.Q4_K_M.gguf", n_gpu_layers=99)
ctx = Context(model, n_ctx=4096)
print(ctx.generate("Hello, AI!", 256))

What you can build with DeepSeek Coder V2

  • code
  • completion
  • code agents

Hardware it fits on

  • Apple Silicon (8GB): 16B in MLX
  • Apple Silicon (16GB): 236B in MLX
  • Apple Silicon (32GB): 236B in MLX or GGUF
  • Apple Silicon (64GB+): 236B in MLX
  • RTX 3090 (24GB): 236B at Q4_K_M
  • RTX 4090 (24GB): same as 3090, ~30% faster
  • RTX 5090 (32GB): 236B at Q8_0
  • 2× RTX 6000 Ada (96GB total): 236B at Q4_K_M

Related models in the DeepSeek family

  • DeepSeek R1 — 1.5B, 7B, 8B, 14B, 32B, 70B, 671B · reasoning, math, step-by-step problem solving
  • DeepSeek V3 — 671B (37B active MoE) · frontier-class open-weights LLM (needs data-center GPUs)
  • DeepSeek Coder — 1.3B, 6.7B, 33B · small-footprint code completion

Other models you might consider

  • Llama 3.1 (Meta) — 8B, general-purpose chat, agentic workflows, code
  • Llama 3.2 (Meta) — 1B, low-end hardware, mobile, edge deployment
  • Llama 3.2 Vision (Meta) — 11B, vision + chat on desktop GPUs
  • Llama 3.3 (Meta) — 70B, best-in-class 70B for desktops
  • Llama 4 (Meta) — Scout 17B (109B MoE), vision + long context (10M tokens)
  • Qwen 2.5 (Alibaba) — 0.5B, strongest 7B-32B on consumer hardware