DeepSeek

Global

DeepSeek V4 API — the cheapest frontier-quality pricing on the market. V4 Flash starts at $0.14 per 1M input tokens with a 98% cache-hit discount, and both tiers offer a 1M context window with open-source weights under the MIT license.

Verified 2026-09-26

Visit

DeepSeek offers the most cost-effective AI API pricing with its V4 series models. With a 128K context window and aggressive pricing, it is ideal for high-volume token consumption and budget-conscious development.

DeepSeek has rapidly emerged as a disruptive force in the AI pricing landscape. Their V4 Flash and V4 Pro models deliver competitive performance at a fraction of the cost of US-based providers. This makes DeepSeek an excellent choice for applications with heavy token consumption where cost efficiency is paramount.

The 128K context window provides ample space for processing medium-to-large documents and conversations. DeepSeek's API platform supports both chat completions and code generation with response times that rival premium providers.

For startups, research projects, and high-volume applications, DeepSeek's pricing model is transformative. The platform supports standard OpenAI-compatible API formats, making migration straightforward for existing applications. While model quality may not match GPT-5 or Claude 4.6 on every benchmark, the value proposition for cost-sensitive deployments is unmatched.

DeepSeek Plans

$0.14USD/1M input tokens$0.28 / 1M output tokens$0.14/M

DeepSeek V4 Flash

DeepSeek V4 Flash1M
  • ✓Output $0.28 per 1M tokens
  • ✓Cache hit $0.0028 per 1M tokens
  • ✓98% cache-hit discount — the cheapest cache pricing in the industry
  • ✓1M context window
  • ✓OpenAI-compatible API, drop-in replacement
  • ✓Open-source weights (MIT), self-hostable
$0.435USD/1M input tokens (75% promotional discount)$0.87 / 1M output tokens$0.43/M

DeepSeek V4 Pro

DeepSeek V4 Pro1M
  • ✓List price $1.74 / $3.48 — currently at 75% off
  • ✓Output $0.87 per 1M tokens
  • ✓Cache hit $0.003625 per 1M tokens
  • ✓1M context window
  • ✓Best value reasoning tier; within ~3% of GPT-5.4 on quality benchmarks
  • ✓Open-source weights (MIT), self-hostable
Free

DeepSeek App & Web

DeepSeek V4
  • ✓Free chat on web and mobile
  • ✓1M context window in the app
  • ✓No credit card required

Other Token Plans