DeepSeek pricing - 2026
DeepSeek API starts at $0.140/1M input tokens. Prices effective since April 30, 2026.
| Model | Input $/1M | Output $/1M | Cached | Notes |
|---|---|---|---|---|
| Deepseek Chat | $0.280 | $0.420 | $0.0280 | |
| Deepseek Coder | $0.140 | $0.280 | — | |
| Deepseek R1 | $0.550 | $2.19 | — | |
| Deepseek Reasoner | $0.280 | $0.420 | $0.0280 | |
| Deepseek V3 | $0.270 | $1.10 | $0.0700 | |
| Deepseek V3.2 | $0.280 | $0.400 | — |
Prices effective since April 30, 2026. Verified May 27, 2026. Confirm at LiteLLM before billing.
Cost calculator
Estimated monthly cost · 70% input / 30% output split
Price history
Input price per 1M tokens - tracked from Apr 30, 2026
Prices scraped daily from official provider documentation. Chart shows input token pricing.
DeepSeek usage limits by plan
| Context Window | 64000 tokens | DeepSeek-V3 and R1 |
| Messages Per Day | 50 messages/day | Web chat, resets daily |
| File Upload Size | 20 MB | Per file |
| Concurrent Requests | 1 requests | Web interface |
| Rpm V3 | 60 requests/min | DeepSeek-V3 |
| Rpm R1 | 60 requests/min | DeepSeek-R1 |
| Tpm V3 | 60000 tokens/min | DeepSeek-V3 |
| Tpm R1 | 60000 tokens/min | DeepSeek-R1 |
| Context Window | 64000 tokens | Both models |
DeepSeek features and capabilities
Generation
| Image generation | ✕ No | DeepSeek does not support image generation. | |
| Web search | ✕ No | No real-time web browsing; knowledge cutoff applies. | |
| Code generation | ✓ Yes | Strong coding performance, especially on competitive benchmarks. | |
| Reasoning mode | ✓ Yes | DeepSeek-R1 uses chain-of-thought reasoning. |
Input & Context
| Audio input | ✕ No | No audio file support. | |
| Video input | ✕ No | No video support. | |
| Image input | ✓ Yes | Multimodal image understanding on latest models. | |
| File upload | ✓ Yes | PDF and document upload supported. |
Integrations & API
| API access | ✓ Yes | OpenAI-compatible REST API. | |
| Plugins/tools | ✕ No | No plugin ecosystem currently. |
Languages
| Chinese | ✓ Yes | Native Chinese language support. | |
| English | ✓ Yes | Full English support. |
Memory
| Memory | ✕ No | No persistent memory across sessions. | |
| Custom instructions | ✓ Yes | System prompt supported via API. |
Privacy & Security
| Training opt-out | ✕ No | Data may be used for model training. | |
| GDPR compliant | ✕ No | Servers in China; limited GDPR coverage. |
Related pages
About DeepSeek API pricing
DeepSeek API pricing is set by DeepSeek and billed per million tokens processed. Input tokens (your prompt) and output tokens (the response) are priced separately. Cached input pricing applies when the same context is reused across requests, offering significant savings for repeated prompts.
Prices on this page are sourced from official DeepSeek documentation and updated when DeepSeek announces pricing changes. Check the official DeepSeek pricing page for the most current rates.
Weekly AI pricing & uptime digest
Price drops, new model releases, and incident summaries - every Monday. Free.