DATASHEET
DeepSeek/RELEASED Sep 10, 2026
DeepSeek V4.1 Flash
NEW ARCHITECTURE FLAGSHIPReleased Sept 10, 2026; smallest model in DeepSeek's new Causal Encoder-Decoder architecture family with native image+text input, 1M context and 384K max output. Outperforms V4-Pro across performance, cost and speed; V4-Flash models retired and V4-Pro routes to V4.1-Flash from Sept 14. Peak $0.3/M in / $1.2/M out, 50% off off-peak.
Context
1,000,000 tokens
Architecture
552B MoE (8B / 16B Activated)
Official API (1M)
$0.3 in / $1.2 out (Open)+ Weights Free to Self-Host
License
MIT License
Verified Benchmark Suite
SWE-BENCH
74.2% (DeepSWE v1.1)
MMLU-PRO
74.1%
GPQA DIAMOND
90.9%
Quickstart Snippet
curl -s https://modelregistry.tirup.in/api/cli?model=deepseek-v4-1-flash