DeepSeek V4.1 Flash and V4 Pro are live

DeepSeek tokens, dirt cheap.

A flat monthly plan, starting at $5, gets you hundreds of millions of DeepSeek tokens. The API is OpenAI-compatible, so you change your base URL and nothing else.

$5to start, billed monthly
100M+tokens on the smallest plan
1Mcontext window
1 lineof code to switch

Drop it into the code you already have.

Point any OpenAI SDK, agent or IDE plugin at api.cheapseek.ai/v1 and your existing code keeps working.

  • Streaming, tool calls and JSON mode
  • Reasoning tokens reported in usage
  • Automatic prompt caching, so cached input counts for less
  • Requests keep going at pay-as-you-go rates once your plan runs out
from openai import OpenAI

client = OpenAI(
    base_url="https://api.cheapseek.ai/v1",
    api_key="cs-...",
)

stream = client.chat.completions.create(
    model="deepseek-flash",
    messages=[{"role": "user", "content": "Refactor this function"}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

Pick a plan. Use a lot of tokens.

Each plan comes with a monthly allowance and a rolling 7-day cap set at ¼ of that allowance, so no single week can drain the whole month. You can upgrade or cancel whenever you like.

Usage factors: deepseek-flash ×1 and deepseek-v4-pro ×3, so a Pro token uses three Flash tokens of allowance. A cached input token counts as 0.1. Once you pass a cap, requests continue at pay-as-you-go rates.

What a heavy month costs.

Take a typical agentic-coding month of 250M input tokens (mostly cache hits) and 25M output tokens on DeepSeek V4.1 Flash.

OptionApprox. monthly cost
Typical closed frontier API$400+
DeepSeek pay-as-you-go, peak hours~$40
CheapSeek Builder$10 flat

Works with your tools.

If a tool lets you set an OpenAI-compatible base URL, it works with CheapSeek.

OpenAI SDKClineRoo CodeOpenCodeContinueAiderLangChainLlamaIndexVercel AI SDKn8n

Questions

Is this the real DeepSeek model?

Yes. We serve the official open-weight DeepSeek models at full precision, and every response includes the exact model version it came from.

What happens when I run out?

Your requests don't stop. Usage over the cap is billed at pay-as-you-go rates from your prepaid balance. If you'd rather stop at the cap, turn on hard limits in the dashboard.

Why is there a weekly cap?

It keeps prices low for everyone. A few accounts can't use a whole month of capacity in a single afternoon. The window rolls, so usage drops out of it seven days after it happened.

Do you train on my data?

No. We don't store prompts or completions beyond what billing and abuse prevention require.

Can I upgrade mid-month?

Yes. When you upgrade, the new plan starts right away with a fresh allowance, and we prorate what you hadn't used on the old one.

How do I pay?

Card through Stripe. Subscriptions renew monthly or yearly, and you can cancel from the billing page whenever you want.

Stop rationing tokens.

Have a key and be streaming in under a minute.