DeepSeek · Updated 2026-07-22

DeepSeek-R1 API Pricing 2026: $0.55/M Tokens

Complete per-million-token pricing for DeepSeek-R1 from DeepSeek, including cached and batch rates, cost-per-1K/1M/10M breakdowns, a monthly cost calculator, and a head-to-head comparison with o3, Claude Sonnet 4, Gemini 2.5 Pro. For full cross-provider rankings, open the main calculator.

DeepSeek-R1 headline pricing

Input

$0.55 /M tokens

Output

$2.19 /M tokens

Context · capabilities

128K context · Code, Reasoning · benchmark 8.5/10

DeepSeek-R1 cost per 1K, 1M and 10M tokens

Standard on-demand USD rates. Input assumes prompt tokens; output assumes generated tokens.

VolumeInput costOutput costTotal (1:1)
1K tokens$0.0006$0.0022$0.0027
1M tokens$0.55$2.19$2.74
10M tokens$5.50$21.90$27.40

DeepSeek-R1 monthly cost calculator

1M
500K

$1.65

per month for 1M in + 500K out

$19.74

per year

Need to weigh it against other models? Use the full cross-provider calculator.

DeepSeek-R1 vs o3 vs Claude Sonnet 4 vs Gemini 2.5 Pro

DeepSeek-R1 ranks #1 of 4 by blended per-million-token price in this comparison. Lower input + output is cheaper.

ModelInput $/MOutput $/MIn+OutContext
YOUDeepSeek-R1DeepSeek$0.55$2.19$2.74128K
Gemini 2.5 ProGoogle$1.25$10.00$11.251M
Claude Sonnet 4Anthropic$3.00$15.00$18.00200K
o3OpenAI$10.00$40.00$50.00200K

Related pricing pages

DeepSeek-R1 pricing FAQ

How much does DeepSeek-R1 cost?

DeepSeek-R1 is priced at $0.55 per million input tokens and $2.19 per million output tokens on DeepSeek. At a typical 2:1 input-to-output ratio, that works out to roughly $1.10 per million blended tokens. Context window is 128K.

Is DeepSeek-R1 cheaper than o3?

Yes — DeepSeek-R1 is cheaper. Its blended input+output rate ($2.74/M) undercuts o3 ($50.00/M) by about 95%. On input alone DeepSeek-R1 is $0.55/M vs $10.00/M; on output $2.19/M vs $40.00/M.

Does DeepSeek-R1 support prompt caching?

No. DeepSeek-R1 does not currently publish a discounted cached-input rate, so all input is billed at the standard $0.55/M. If your workload is cache-heavy, consider o3 or another model that supports prompt caching.

Can I use DeepSeek-R1 with the Batch API?

No published batch discount is available for DeepSeek-R1. For non-latency-sensitive workloads, compare against o3 or another model that offers a batch tier.

☕ Support this tool