DeepSeek · open · 2026-04

DeepSeek-V4-Flash

#10 index 85.2
the take

The same trick as V4-Pro in a size you could almost self-host, for pennies. Same caveat too: the scores are DeepSeek's own until someone neutral runs them.

benchmarks
specs
Context
1M
Input
$0.14/M
Output
$0.28/M
Speed
Modality
text
strengths
  • Among the cheapest frontier-tier APIs
  • 1M context in a compact MoE
  • Open weights, strong reported coding
weaknesses
  • Vendor-reported, unverified scores
  • Trails V4-Pro on the hardest tasks
  • Text-only

Sources: [1][2]