DeepSeek API pricing
Para equipos técnicosDeepSeek is priced per use rather than per month; the rate for each option is in the table below. Checked 11 sept 2026 against the official pricing page.
Modelos de frontera muy baratos. V4 Flash ($0.14/$0.28 por 1M) es la opción de volumen; V4 Pro-0813 son pesos nuevos en el mismo id de API a $0.435/$0.87 — no un relanzamiento de precio.
DeepSeek is still one of the lowest-cost ways to run a capable model through an API, with both OpenAI- and Anthropic-compatible endpoints. Pricing now depends on the time of day, so batch work you can schedule off-peak costs half as much.
Plans and prices
| Plan | Price | What you get |
|---|---|---|
| deepseek-flashOur pick | Usage-basedDeepSeek-V4.1-Flash, effective 04:00 UTC 2026-09-10. Input cache-miss $0.15/1M off-peak, $0.30 peak; input cache-hit $0.003/$0.006; output $0.60/$1.20. Peak 01:00-04:00 & 06:00-10:00 UTC Mon-Fri; off-peak is 50% of peak. The legacy ids deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted but route here at these rates. | 1M context, 384K max output, vision supported, concurrency limit 2500
|
| deepseek-v4-pro | Usage-basedNOT retired after all. DeepSeek announced on 2026-09-10 that from 04:00 UTC on 2026-09-14 all deepseek-v4-pro requests would route to V4.1 Flash, then reversed it: DeepSeek states it will continue providing API services for DeepSeek V4 Pro after September 14, 2026, with the billing method remaining unchanged. Re-confirmed on the live pricing page 2026-09-18, four days after the announced cutover. Rates (model version DeepSeek-V4-Pro-0813): input cache-miss $0.66/1M off-peak, $1.32 peak; cache-hit $0.022/$0.044; output $1.98/$3.96. Peak 01:00-04:00 & 06:00-10:00 UTC Mon-Fri. Since V4.1 Flash now beats it on DeepSeek own agentic benchmarks at about a quarter of the price, V4 Pro is a fallback rather than a default step-up. | 1M context, concurrency limit 500, no vision
|
Pick this tier: deepseek-flash
The Flash model is the cheapest DeepSeek option and covers most workloads. Schedule batch jobs off-peak to halve the bill.
What works
- Very low per-token prices compared with the frontier labs
- Off-peak hours bill at half the peak rate
- Cache hits cost a small fraction of fresh input
- Drop-in compatible with OpenAI and Anthropic API clients
What to watch
- Peak and off-peak billing makes cost forecasting harder
- Model names and routing change quickly; check which model an ID now serves
- Data-location requirements rule it out for some businesses
Skip it if
- Your data must stay with a provider in your own region
- You need a chat subscription rather than an API
- You cannot tolerate model IDs being retired and rerouted at short notice
Ideal para
DeepSeek pricing questions
- How much does DeepSeek cost in 2026?
- DeepSeek is priced per use rather than per month; the rate for each option is in the table below. Checked 11 sept 2026 against the official pricing page.
- Does DeepSeek API pricing change by time of day?
- Yes. DeepSeek bills a peak rate and an off-peak rate, and off-peak is half of peak. Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays; every other hour bills off-peak.
- What happened to deepseek-v4-flash?
- DeepSeek retired the deepseek-v4-flash model ID on 2026-09-10. Calls to it now route to V4.1 Flash and bill at the V4.1 Flash rates.
- Is DeepSeek's API compatible with OpenAI and Anthropic clients?
- Yes. DeepSeek offers OpenAI-compatible and Anthropic-compatible endpoints, so most existing client libraries work by changing the base URL and key.
- How do I pay less for the DeepSeek API?
- Run non-urgent jobs in off-peak hours, which bill at half the peak rate, and reuse prompt prefixes so input is billed at the much lower cache-hit rate.
Reviewed 11 sept 2026 by the Dapols team. Report a price change
Nuestra evidencia sobre esta herramienta
- Estado en el catálogo
- Recomendada: puede ser una elección principal en un plan
- Precio comprobado por última vez
- 11 sept 2026
- Capacidad revisada por última vez
- 11 sept 2026
- Confianza de la evidencia
- Media: verificada, pero el proveedor cambia los planes a menudo
- Documentación oficial
- Leer la documentación del proveedor
Verificado nuevamente en lotes. Fechamos cada verificación y nunca aceptamos pago por la ubicación. Cómo clasificamos las herramientas
Is DeepSeek right for your business?
Answer a few questions and see whether DeepSeek fits your stack and budget.