DeepSeek V4.1 Flash
DeepSeek · Multimodal · Released Sep 2026
A sparse mixture-of-experts model from DeepSeek with 13 billion active parameters out of 284 billion total and a 1-million-token context window, designed as the cost-efficient tier of the V4.1 family.
- Strengths
- Delivers higher performance and speed than V4 Pro while maintaining lower inference costs through its sparse architecture.
- Best for
- Production workloads where throughput and latency matter as much as capability, and where the reduced parameter activation yields meaningful cost savings.
- Limitations
- As the smaller, faster model in the V4.1 lineup, it trades some reasoning depth and capability ceiling for efficiency; tasks requiring frontier-class performance may benefit from V4 Pro instead.
Input / 1M
$0.15
Output / 1M
$0.6
Cached input / 1M
$0.003
Context window
1.05M
Price history
Input (solid)Output (dashed)
Price change
- 30d
- in decreased 0.0% out decreased 0.0%
- 90d
- in decreased 0.0% out decreased 0.0%
- 1y
- in decreased 0.0% out decreased 0.0%
- Since launch
- in decreased 0.0% out decreased 0.0%
Snapshots
| Effective | Input | Output | Cached in | Note | Source |
|---|---|---|---|---|---|
| 11 Sep 2026 | $0.15 | $0.6 | $0.003 | Imported from OpenRouter | openrouter.ai |
| 10 Sep 2026 | $0.15 | $0.6 | $0.003 | Imported from OpenRouter | openrouter.ai |