DeepSeek API Cost Calculator 2026
DeepSeek API Cost Calculator
Estimate monthly API costs for DeepSeek V3 and R1 with prompt caching
Prompt caching
90% off input price for cached tokens. Most effective for repeated system prompts or long shared context.
Cost per request
$0.05
DeepSeek V3 · 100K input / 20K output
Monthly cost
$49.00
1,000 requests/mo · Annual: $588.00
Cost breakdown per request
Model comparison
Same token settings applied to all models. DeepSeek models reflect current cache settings.
| Model | Input/MTok | Output/MTok | Per request | Monthly |
|---|---|---|---|---|
GPT-4o mini | $0.15 | $0.6 | $0.03 | $27.00 |
Gemini 2.5 Flash | $0.15 | $0.6 | $0.03 | $27.00 |
DeepSeek V3Best value vs OpenAISelected | $0.27 | $1.1 | $0.05 | $49.00 |
DeepSeek V3 0324 | $0.27 | $1.1 | $0.05 | $49.00 |
DeepSeek R1 | $0.55 | $2.19 | $0.10 | $98.80 |
Claude Haiku 4.5 | $0.8 | $4 | $0.16 | $160.00 |
Prices per million tokens as of June 2026. Verify at provider pricing page before use -- rates change frequently.
Quick Answer
How much does the DeepSeek API cost?
DeepSeek V3 costs $0.27 per million input tokens and $1.10 per million output tokens as of June 2026. This makes it 9x cheaper than GPT-4o on input tokens. DeepSeek R1 (reasoning model) costs $0.55/$2.19 per million tokens. Both models support prompt caching at 90 percent off input price.
About this tool
This calculator estimates monthly DeepSeek API costs based on your token usage and request volume. Enter your expected input and output tokens per request, set your monthly request count, and optionally enable prompt caching to see how much you can save. The results update in real time and include a comparison against GPT-4o mini, Claude Haiku 4.5, and Gemini 2.5 Flash for the same token counts.
DeepSeek V3 and R1 represent a new tier of cost-effective frontier models. At $0.27 per million input tokens, DeepSeek V3 is roughly 9x cheaper than GPT-4o and about 3x cheaper than Claude Haiku 4.5. For teams running high-volume inference workloads or building cost-sensitive applications, the difference can be significant. Use this tool to model your costs before committing to a provider. Verify at provider pricing page before use -- rates change frequently.
How it works
DeepSeek API pricing (June 2026)
| Model | Input | Output | Cached input | Best for |
|---|---|---|---|---|
| DeepSeek V3 | $0.27/MTok | $1.10/MTok | $0.027/MTok | General tasks, content, chat |
| DeepSeek R1 | $0.55/MTok | $2.19/MTok | $0.055/MTok | Reasoning, coding, math |
| DeepSeek V3 0324 | $0.27/MTok | $1.10/MTok | $0.027/MTok | Latest V3, improved coding |
Prices per million tokens. Cached input is 90% off standard input price.
DeepSeek vs competitors
Monthly cost at 1M input tokens/day + 200K output tokens/day (30M input, 6M output per month). No caching applied.
| Model | Input/MTok | Output/MTok | Monthly cost |
|---|---|---|---|
| Gemini 2.5 Flash | $0.15 | $0.60 | $8.10 |
| GPT-4o mini | $0.15 | $0.60 | $8.10 |
| DeepSeek V39x cheaper than GPT-4o | $0.27 | $1.10 | $14.70 |
| Claude Haiku 4.5 | $0.80 | $4.00 | $48.00 |
Based on published API pricing June 2026. Use the calculator above for your specific token counts.
When to use DeepSeek vs OpenAI or Anthropic
Use DeepSeek when
Cost is the primary factor and you can tolerate variable availability
- ✓You need to cut API costs by 70 to 90 percent vs GPT-4o
- ✓Building personal projects, startups, or cost-sensitive applications
- ✓Running high-volume batch processing where timing is flexible
- ✓Complex reasoning or coding tasks where R1 excels at low cost
- ✓Experimenting with long-context workloads before committing to a provider
Use OpenAI or Anthropic when
Reliability, compliance, and ecosystem matter more than lowest price
- ✓You require guaranteed uptime and enterprise SLA commitments
- ✓Your use case needs strong data privacy or compliance agreements
- ✓Working with multimodal inputs including images, audio, or files
- ✓Your team is already integrated with OpenAI or Anthropic tooling
- ✓You need dedicated customer support with guaranteed response times
Frequently asked questions about DeepSeek API pricing
How much does DeepSeek API cost?+
DeepSeek V3 costs $0.27 per million input tokens and $1.10 per million output tokens as of June 2026. DeepSeek R1 costs $0.55 input and $2.19 output per million tokens. Both models support prompt caching at 90 percent off input price.
Is DeepSeek cheaper than GPT-4o?+
Yes. DeepSeek V3 at $0.27/$1.10 per million tokens is roughly 9x cheaper on input and 9x cheaper on output than GPT-4o at $2.50/$10.00. For high-volume applications, DeepSeek can reduce API costs by 80 to 90 percent compared to GPT-4o.
What is DeepSeek R1 used for?+
DeepSeek R1 is a reasoning model optimized for complex tasks like coding, math, and multi-step logical reasoning. It costs $0.55/$2.19 per million tokens, roughly double V3, but outperforms GPT-4o on many reasoning benchmarks at a fraction of the cost.
Does DeepSeek support prompt caching?+
Yes. DeepSeek V3 and R1 both support prompt caching at 90 percent off the standard input price. Cached tokens cost $0.027 per million for V3 and $0.055 per million for R1. Caching is effective for applications with repeated system prompts or context.
How does DeepSeek pricing compare to Claude Haiku?+
DeepSeek V3 at $0.27/$1.10 is cheaper than Claude Haiku 4.5 at $0.80/$4.00 per million tokens. For pure cost, DeepSeek V3 wins. However, Claude Haiku has stronger instruction following and is hosted by Anthropic with enterprise SLAs.
What is the difference between DeepSeek V3 and V3 0324?+
DeepSeek V3 0324 is an updated checkpoint of the original V3 model released in March 2024, with improved coding and instruction following performance. Both share the same pricing at $0.27/$1.10 per million tokens.
Can I use DeepSeek API for production applications?+
DeepSeek API is suitable for production use for many applications, but availability and rate limits may be more variable than OpenAI or Anthropic. For mission-critical applications requiring guaranteed uptime, enterprise support, or data privacy agreements, evaluate the trade-offs carefully.
How do I reduce DeepSeek API costs further?+
Enable prompt caching for repeated context to save 90 percent on cached input tokens, batch similar requests together, use DeepSeek V3 for most tasks and reserve R1 only for complex reasoning, and monitor token usage with the DeepSeek dashboard to identify optimization opportunities.