DeepSeek V3.1
DeepSeek · Released Aug 2025
DeepSeek V3.1 is a 671-billion-parameter hybrid reasoning model with 37 billion active parameters that supports both thinking and non-thinking inference modes via prompt templates.
- Strengths
- Handles long-context work with a 163,840 token window and can apply extended reasoning when needed through its two-phase architecture.
- Best for
- Complex problem-solving tasks where reasoning steps are valuable, and applications requiring processing of lengthy documents or conversations in a single request.
- Limitations
- The active-parameter architecture and reasoning modes add latency compared to non-reasoning models, and thinking mode consumes additional tokens without being directly visible in output.
Input / 1M
$0.25
Output / 1M
$0.95
Cached input / 1M
$0.13
Context window
163K
Price history
Input (solid)Output (dashed)
Price change
- 30d
- in increased 19.0% out increased 20.3%
- 90d
- in increased 19.0% out increased 20.3%
- 1y
- in increased 19.0% out increased 20.3%
- Since launch
- in increased 19.0% out increased 20.3%
Snapshots
| Effective | Input | Output | Cached in | Note | Source |
|---|---|---|---|---|---|
| 13 Jul 2026 | $0.25 | $0.95 | $0.13 | Imported from OpenRouter | openrouter.ai |
| 11 Jun 2026 | $0.21 | $0.79 | $0.13 | Imported from OpenRouter | openrouter.ai |