The Open-Source AI Stack
RSS
All models

Models · deepseek

DeepSeek-V2 Chat

Open DeepSeek · 2024-05-07 · DeepSeek License

The Multi-head Latent Attention (MLA) debut paper. Cut KV-cache memory by ~93% versus dense attention, making 128K-context MoE inference economically viable.

Cost

— / Mtok input
— / Mtok output

DeepSeek API · as of 2026-05-19

via Artificial Analysis ↗

Speed

— tok/sec output

DeepSeek API · as of 2026-05-19

via Artificial Analysis ↗

Architecture

tokens in Embedding vocab not disclosed · deepseek tokenizer × N layers Multi-head Latent Attention RoPE + YaRN context 128,000 tokens MoE Router 160 experts total · 6 active per token shown: 32 of 160 Output projection tokens out
Schema-generated from data/models.yaml. Every label is auditable against the model's sources.

Specs

Architecture
moe
Total params
236B
Active params
21B
Experts
160 total · 6 active
Context window
128K tokens
Attention
mla
Position encoding
rope-yarn
Pretraining tokens
8.1T
Post-training
sft, rlhf
OSI-approved
no
Data released
no
Training code
not released

Benchmarks

Each score carries the date it was published; we never infer or interpolate missing scores.

Recommended use cases

  • cost-efficient chat at MoE economics
  • long-context retrieval

Available quantizations

GGUF llama.cpp's container; the common local format, k-quants from Q2 to Q8. runs on llama.cpp, Ollama
MLX Apple MLX 4/8-bit layout for Apple silicon. runs on Apple MLX

Verified via the Hugging Face model tree ↗. Community quantizations change over time; the families shown are those with published weights at audit time.

Notable innovations

  • · Multi-head Latent Attention (MLA)
  • · KV-cache compression

Lineage

MLA debut; architecture refined into V3.

Derivatives

Sources