Deepseek Plans Pricing
DeepSeek prices its API on a per-token, pay-as-you-go basis with no monthly subscription tier. Two production models are available — DeepSeek-V4-Flash (non-thinking) and DeepSeek-V4-Pro (thinking/reasoning). Both expose distinct prices for cache-hit and cache-miss input tokens plus output tokens. The legacy `deepseek-chat` and `deepseek-reasoner` model identifiers are deprecated aliases. V4-Pro carries a 75 percent promotional discount valid through 2026-05-31. Account top-ups are prepaid; there is no separate enterprise SKU on the public site.
Plans
Pay-as-you-go pricing for DeepSeek-V4-Flash (non-thinking mode), the default chat and FIM completion model. Replaces the deprecated `deepseek-chat` alias.
- 1M token context window
- 384K max output tokens
- Chat Completions API
- Fill-In-The-Middle (FIM) Completions API
- JSON output
- Tool calls
- Chat prefix completion
- Context caching
Pay-as-you-go pricing for DeepSeek-V4-Pro (thinking / reasoning mode). Replaces the deprecated `deepseek-reasoner` alias. Listed prices reflect the 75 percent promotional discount in effect through 2026-05-31.
- 1M token context window
- 384K max output tokens
- Reasoning / thinking mode
- Chat Completions API
- JSON output
- Tool calls
- Chat prefix completion
- Context caching
- 75% promotional discount through 2026-05-31