/ Models Index / DeepSeek V3.5

DeepSeek V3.5

Open-Weights · DeepSeek · DeepSeek · Released Q1 2026
A-

Excellent (A-) — Capability Grade

DeepSeek V3.5 is DeepSeek's current-generation flagship — multimodal (text + vision) added, longer context, improved reasoning. MIT licensed remains a category differentiator. Strong on-premise deployment economics.

87
Composite / 100
/ Subscore Breakdown · 6 Capability Dimensions

Where this grade comes from.

General Reasoning
A-
Code Generation
A-
Math & STEM
A-
Tool Use & Agency
B+
Multimodal
B
Safety & Alignment
B+
/ Key Events & Disclosures

Release timeline & positioning.

  • Released Q1 2026
  • Multimodal addition
  • MIT licensed
  • Strong on-premise economics

/ Best for

Open-weights frontier-tier deployments; sovereign-AI and on-premise enterprise deployments.

/ Watch out for

China-headquartered may produce regulatory considerations for some enterprises and government deployments.