Model comparison
DeepSeek V3 0324 vs DeepSeek R1
Capability radar, value bars, and side-by-side posture.
DeepSeek V3 0324
Not scored
value score
DeepSeek R1
90
value score
DeepSeek V3 0324 ctx
—
DeepSeek R1 ctx
128K
Capability radar
Value · context · multimodal · openness · speed posture
Value head-to-head
- DeepSeek R190
Attribute tape
| Field | DeepSeek V3 0324 | DeepSeek R1 |
|---|---|---|
| Developer | ~deepseek | ~deepseek |
| Context | — | 128K |
| Modalities | Text, Code | |
| Openness | — | Open Weights |
| Speed | — | — |
| Price | — | — |
| Value score | — | 90 |
| Summary | DeepSeek V3 0324 is a large Mixture-of-Experts (MoE) language model with 671B total parameters (37B active per token), released by DeepSeek in March 2025. It targets general-purpose chat, code generation, long-context reasoning, and tool use, and is published as an open-weight model. | Reasoning-focused open-weights DeepSeek model. |