Wednesday, October 28, 2026 🏢 AI Companies Hub RSS About Contact Admin
POPULAR BEATS: Generative AI LLMs & NLP Autonomous Agents Robotics & Hardware Enterprise AI AI Ethics & Policy 🏢 All AI Companies

DeepSeek-V3

MLA + Auxiliary-loss-free MoE
Official Source & Paper
DeepSeek logo DeepSeek 📄 Open Weights (MIT) 🌍 China (China) 🗓️ Effective Date: 26.12.2024

Landmark 671B parameter Mixture-of-Experts base model demonstrating frontier capability trained on only $6M in compute.

Model Size / Parameters 671B (37B active) Architecture: MLA + Auxiliary-loss-free MoE
Context Window (Input) 128k tokens Token Capacity
Max Output Tokens 8,192 tokens Single Generation Limit
Supported Modalities Text, Code Native Multimodal Ingestion
License & Access Open Weights (MIT) Open Weights vs. API
Release Date December 2024 Effective Date: 26.12.2024

Verified Benchmark Results

Standardized academic and industry evaluation benchmarks (Effective Date: 26.12.2024).

MMLU-Pro / 5-Shot General Intellect
88.5%
SWE-bench Verified Software Engineering
42.0%
MATH 500 / AIME Mathematical Reasoning
89.3%
GPQA Diamond Doctoral Science
65.9%
HumanEval / Python Algorithm Synthesis
82.6%

Pricing & Cost Efficiency

Official API token pricing and self-hosted licensing terms.

Input Pricing (Prompt)
$0.14 / 1M tokens
Per 1 million processed tokens
Output Pricing (Generation)
$0.28 / 1M tokens
Per 1 million generated tokens
Verified Datasheet: Source: DeepSeek Research Documentation • Effective Date: 26.12.2024
← Back to All AI Model Datasheets

Latest Coverage & Analysis on DeepSeek-V3