DeepSeek V4 Flash ranks #2 open-weight model at 50x lower cost
Priced at $0.09 per 1M tokens, DeepSeek's model challenges premium AI economics.
Deep Dive
DeepSeek's new flash model is unexpectedly cheap and high-performing across useful benchmarks, priced at $0.09 / $0.18 per 1M tokens. The poster calls it "intelligence too cheap to meter" and says it works pretty well on coding and reasoning chat topics.
Key Points
- DeepSeek V4 Flash is ranked ~#2 among open-weight models, trailing only Kimi K3 on useful benchmarks.
- Priced at $0.09 input / $0.18 output per 1M tokens, making it over 50x cheaper than leading proprietary models.
- Early testing shows strong performance on coding and reasoning tasks, driving interest for production AI applications.
Why It Matters
Cheap open-weight models like V4 Flash slash AI inference costs, unlocking large-scale agentic and coding use cases.