DeepSeek: 50x Cheaper AI Tokens Than OpenAI

DeepSeek now delivers AI token inference up to 50x cheaper than OpenAI and Anthropic after its permanent 75% price reduction that dropped select costs to fractions of a cent per million tokens. The move builds on the team’s self-developed training framework and self-built compute clusters that enabled rapid release of open-source models including DeepSeek-LLM and DeepSeek-Coder within six months. Input, output and cached token pricing structures continue to separate DeepSeek from rivals on 2025-2026 AI token inference cost benchmarks while deepening pressure across the inference market.