DeepSeek V4.1-Flash: 552B MoE, Native Vision, New Pricing
Overview
DeepSeek-V4.1-Flash launched September 10, 2026 as the successor to the V4-Flash line — the newest fast-tier model on DeepSeek's API (call name deepseek-flash).
Key Specs
- 552B-parameter MoE with a restructured encoder design
- Native multimodal: accepts images directly without external vision modules
- Performance approaches or exceeds V4 Pro (0813) on multiple benchmarks
API Pricing (per 1M tokens)
| Item | Off-peak | Peak |
|---|---|---|
| Input | $0.15 | $0.30 |
| Output | $0.60 | $1.20 |
| Cache-hit input | $0.003 | |
Output pricing is roughly 70% lower than the previous flagship tier. Since September 14, 2026, v4-pro API calls route to V4.1-Flash.