DeepSeek V3 0324
DeepSeek · Released Mar 2025
DeepSeek V3 is a 685-billion-parameter mixture-of-experts model designed for multi-turn conversations and reasoning tasks.
- Strengths
- Handles long contexts efficiently with a 163K token window and shows strong performance on reasoning, code generation, and knowledge-heavy tasks.
- Best for
- Complex reasoning, code generation, and applications requiring extended context windows or sustained multi-turn interactions.
- Limitations
- As a very large model, it may have higher latency than smaller alternatives and requires careful prompt engineering to avoid verbose or off-topic outputs.
Input / 1M
$0.27
Output / 1M
$1.12
Cached input / 1M
$0.135
Context window
163K
Price history
Input (solid)Output (dashed)
Price change
- 30d
- in increased 35.0% out increased 45.5%
- 90d
- in increased 35.0% out increased 45.5%
- 1y
- in increased 35.0% out increased 45.5%
- Since launch
- in increased 35.0% out increased 45.5%
Snapshots
| Effective | Input | Output | Cached in | Note | Source |
|---|---|---|---|---|---|
| 15 Jul 2026 | $0.27 | $1.12 | $0.135 | Imported from OpenRouter | openrouter.ai |
| 2 Jul 2026 | $0.24 | $0.9 | $0.135 | Imported from OpenRouter | openrouter.ai |
| 11 Jun 2026 | $0.2 | $0.77 | $0.135 | Imported from OpenRouter | openrouter.ai |