Provider pricing
DeepSeek AI API pricing
DeepSeek offers the best price-to-performance ratio in the industry. The April 2026 V4 launch brought V4 Pro (1.6T-param MoE flagship at $0.55/$2.20 per M tokens) and V4 Flash (284B-param MoE at $0.14/$0.55), both with native prompt caching. V3.2 and R1 remain available for legacy workloads.
Founded2023
HeadquartersHangzhou, China
Tracked models9
Best for
Cost-conscious teams that want flagship-quality output at budget prices.
Strengths
- V4 Pro: frontier-class reasoning at a fraction of GPT-5.5/Claude Opus pricing
- V4 Flash: ~3.5× faster than V4 Pro, ideal for high-throughput chat/RAG
- 90% cache hit discount — industry leading
- Open-weight models available for self-hosting
Tracked DeepSeek models
| Model | Input $/M | Output $/M | Context | Tier | Best for |
|---|---|---|---|---|---|
| DeepSeek V4 Pro | $0.550 | $2.20 | 256,000 | Flagship | 1.6T MoE flagship (49B active), frontier reasoning at open-weight prices |
| DeepSeek V4 Flash | $0.140 | $0.550 | 128,000 | Budget | 284B MoE (13B active), high-throughput chat/RAG/extraction |
| DeepSeek V3.2 | $0.280 | $0.420 | 128,000 | Budget | Best value overall, 90% cache discount |
| DeepSeek R1 | $0.550 | $2.19 | 128,000 | Reasoning | Reasoning at budget prices |
| DeepSeek V3 | $0.270 | $1.10 | 128,000 | Flagship | 671B MoE flagship — strong coding and reasoning at open-weight prices |
| DeepSeek V2.5 | $0.140 | $0.280 | 128,000 | Mid-tier | Previous-gen DeepSeek for general chat and coding |
| DeepSeek Coder V2 | $0.140 | $0.280 | 128,000 | Mid-tier | Code-specialized MoE for IDE completion and repo-scale tasks |
| DeepSeek R2 | $0.700 | $2.80 | 256,000 | Reasoning | Successor to R1 — frontier open-weight reasoning at budget prices |
| DeepSeek Coder V3 | $0.200 | $0.800 | 256,000 | Flagship | Code-specialized MoE for repo-scale completion and refactoring |