DeepSeek-V3

DeepSeek-V3 by deepseek-ai: a 684.5B-parameter mixture-of-experts open-weight model under the unknown license with tool calling.

unknownMixture of experts164K contextTool callingAuto-imported · not yet reviewed

1,131,639 downloads · 4,212 likes · Model card · synced 2026-09-19

Specifications

Parameters: 684.5B (21.4B active)
Architecture: moe · deepseek_v3
Layers: 61
Hidden size: 7,168
KV heads · head dim: 128 · 56
Vocabulary: 129,280
Native dtype: bfloat16
Context window: 163,840 tokens
Released: 2024-12-25
Last modified on Hub: 2025-03-27

Features & licensing

License: unknown
Commercial use: No
Modalities: text
Tool / function calling: Yes
Reasoning mode: No
Languages: EN
Quantised variants on Hub: —
Pipeline: text-generation

transformerssafetensorsdeepseek_v3text-generationconversationalcustom_codeeval-resultstext-generation-inferenceendpoints_compatiblefp8

Task fit

Editorial scores (0–100) used by the recommendation engine.

Chat70
RAG / Q&A70
Code70
Summarisation70
Extraction70
Agents70

Hardware to run DeepSeek-V3

Weights need about 1506 GB at FP16, 791 GB at INT8 and 452 GB at INT4. The KV cache adds roughly 1749 MB per 1,000 tokens per request. For a reference workload of 5 requests per second with 1,500 input and 300 output tokens, total GPU memory is around 1252.4 GB.

GPU configurations at the reference workload
Configuration	VRAM	Utilisation	Est. first token	Cloud / month

Estimates only. Run an assessment for your own traffic.

Auto-imported from Hugging Face. Architecture, license and feature data are synced from the model repository; the task-fit scores below are provisional defaults until a curator reviews them.