▸ AI/LLM · Decode The Future
DeepSeek R2 Launches: 32B Model Rivals Frontier Labs at 70% Lower Cost
DeepSeek released R2, a 32-billion-parameter open-weight reasoning model scoring 92.7% on AIME 2025. The model runs on a single 24GB consumer GPU and undercuts Western frontier APIs by roughly 70% on token cost, pivoting from the rumored 1.2T MoE design.
The bottom line: Marketing data teams can now run production-grade reasoning workloads at a fraction of current LLM costs, fundamentally shifting build-vs-buy economics for analytics agents.