DeepSeek R1 (reasoner) token cost & pricing

DeepSeek R1 exposes chain-of-thought reasoning at $0.55 per million input and $2.19 per million output — roughly one-fortieth of Claude Opus 4.1 on output for a model that can handle competition-grade math and multi-step logic.

DeepSeek R1 (reasoner) specifications

What DeepSeek R1 (reasoner) costs in practice

Four common workloads, priced at list rates. Monthly figures assume 30 days.

How many requests does $100 buy?

Assuming a 1,000-token prompt and a 500-token response per request.

How DeepSeek R1 (reasoner) pricing actually works

The hidden cost is reasoning tokens: the model generates internal thinking that is billed as output. A prompt that produces a 300-token answer may bill 2,000-5,000 output tokens internally, so effective cost per answer runs several times the naive estimate. Budget with that multiplier in mind.

When to use DeepSeek R1 (reasoner)

Use it for math, algorithm design, structured logical deduction and cases where you need to inspect the reasoning trace. Avoid it for simple chat — the reasoning overhead makes it slower and more expensive than a non-reasoning model on easy inputs.

How to cut your DeepSeek R1 (reasoner) bill

Cap the reasoning budget where the API allows it, and route easy queries to DeepSeek chat instead. In practice most pipelines get the best economics by sending everything to chat first and escalating only failures to the reasoner.

Other DeepSeek models

Cheaper alternatives to DeepSeek R1 (reasoner)

• GPT-5 miniOpenAI18% cheaper
• Llama 4 MaverickMeta59% cheaper
• GPT-4o miniOpenAI73% cheaper
• DeepSeek V3.2 (chat)DeepSeek74% cheaper