DeepSeek-V3 vs DeepSeek-V4-Flash
DeepSeek vs DeepSeek — benchmarks, pricing, and capabilities side by side.
- •DeepSeek-V4-Flash has the higher intelligence index (89.4 vs 58.1)
- •DeepSeek-V4-Flash is cheaper ($0.10 vs $0.23 per 1M input)
- •DeepSeek-V4-Flash is faster
- •DeepSeek-V4-Flash has a larger context window (1M)
| DeepSeek-V3 | DeepSeek-V4-Flash | |
|---|---|---|
| Intelligence index | 58.1 | 89.4 |
| Developer | DeepSeek | DeepSeek |
| Type | LLM | LLM |
| Access | Open weights | Open weights |
| Context window | 131,072 tokens | 1,048,576 tokens |
| Input price | $0.23 / 1M | $0.10 / 1M |
| Output price | $0.91 / 1M | $0.20 / 1M |
| Speed | 100 tok/s | 109 tok/s |
| Released | December 26, 2024 | April 24, 2026 |
| Parameters | 671B total / 37B active (MoE) | 284B (13B active) |
| Input modalities | Text | Text |
| Output modalities | Text | Text |
Shared benchmarks
DeepSeek-V3
DeepSeek-V4-Flash
GPQA Diamond
59.1
89.4
Humanity’s Last Exam
3.6
32.1