LLM Cost Calculator

DeepSeek API Pricing 2026

Chinese AI lab delivering frontier open-source models at industry-leading prices. V4 Flash is the default chat tier; R1-0528 is the latest reasoning snapshot with improved performance.

Pricing verified 2026-08-03. Sourced from api-docs.deepseek.com/quick_start/pricing.

Get DeepSeek API access →

DeepSeek Model Pricing

Prices in USD per 1M tokens

ModelInput / 1MOutput / 1MContext
DeepSeek V4 Flash
Default chat model (deepseek-chat alias); cache hit $0.0028/1M input
$0.14$0.281,000,000
DeepSeek V4 Pro
DeepSeek flagship; 1M context; promo 75% off ($0.435/$0.87) until May 31 2026
$0.44$0.871,000,000
DeepSeek V3.2
DeepSeek V3.2; improved over V3 base
$0.27$0.4131,072
R1 0528
DeepSeek R1 May-2025 snapshot; improved reasoning
$0.5$2.15163,840
DeepSeek V3
Legacy V3 pricing; API now routes chat workloads to V4 Flash
$0.26$1.03128,000
DeepSeek V3 (Mar 2025)
March 2025 snapshot; same pricing as V3, improved performance
$0.27$1.1128,000
DeepSeek R1
Reasoning model (deepseek-reasoner alias maps to V4 Flash thinking mode)
$0.7$2.5128,000

How to read these estimates

The monthly table assumes 70% input tokens and 30% output tokens. It is a consistent comparison baseline, not a prediction of your workload. Actual bills can also include cached-input discounts, batch pricing, tool calls, minimum charges, regional differences, retries, and taxes.

The linked source is the provider's pricing documentation. Verify the current rate card and model availability before committing to a budget.

Estimated Monthly Cost (70% input / 30% output split)

Model1M tokens/mo10M tokens/mo100M tokens/mo1B tokens/mo
DeepSeek V4 Flash$0.182$1.82$18.20$182
DeepSeek V4 Pro$0.569$5.69$56.90$569
DeepSeek V3.2$0.309$3.09$30.90$309
R1 0528$0.995$9.95$99.50$995
DeepSeek V3$0.491$4.91$49.10$491
DeepSeek V3 (Mar 2025)$0.519$5.19$51.90$519
DeepSeek R1$1.24$12.40$124$1,240

Frequently Asked Questions

How much does DeepSeek LLM API cost?

DeepSeek offers 7 models ranging from $0.140/1M to $0.70/1M input tokens. Chinese AI lab delivering frontier open-source models at industry-leading prices. V4 Flash is the default chat tier; R1-0528 is the latest reasoning snapshot with improved performance.

Is DeepSeek cheaper than self-hosting?

For low-volume workloads (under 100M tokens/month), cloud APIs like DeepSeek are almost always cheaper than purchasing and maintaining GPU hardware. Use our calculator to find the exact break-even point for your usage.