Home → Guides → DeepSeek API pricing: V4 Pro vs V4 Flash
Two DeepSeek models, two price points. The cheaper one is not always the cheapest bill: V4 Pro thinks before it answers, and every reasoning token is billed as output.
| Model | List input | List output | Here: input | Here: output | Share of list |
|---|---|---|---|---|---|
deepseek-v4-pro | $1.32 | $3.96 | $0.675 | $2.02 | 51.1% |
deepseek-v4-flash | $0.440 | $1.32 | $0.200 | $0.600 | 45.5% |
The figures on this platform are filled in from live billing data each hour, so they always match what you are charged.
On 2 November 2026 we asked both text models to reply with the single word “OK”:
| Model | Output tokens billed | Of which reasoning | Response time |
|---|---|---|---|
deepseek-v4-pro | 23 | 21 | 4.0 s |
deepseek-v4-flash | 12 | 10 | 1.0 s |
Both models think before answering, and the thinking is counted in completion_tokens. For short, simple requests most of the output you pay for is reasoning. Budget by the output column, not by the length of the visible answer, and set max_tokens high enough that the reasoning does not use it all up and leave the reply empty (see empty response with reasoning models).
A coding-agent turn of 100,000 input tokens and 2,000 output tokens, at list price:
deepseek-v4-pro: $0.132 + $0.0079 = $0.140deepseek-v4-flash: $0.044 + $0.0026 = $0.047Input dominates agent workloads, so the input column decides the bill. Multiply by the share of list in the table above for the price here.
deepseek-v4-pro for harder reasoning and agent steps. It does not read images: in our 12 October test it answered “white” for a green image, without an error.deepseek-v4-flash for high-volume, quick work — it answered four times faster in the test above.gpt-6-astra or claude-opus-5-5 (see the image test in our Cherry Studio guide).deepseek-v4-pro called a read_file-style tool with the correct argument names in every run in our Cline tool-call test.
The endpoint is OpenAI-compatible. Base URL https://apiclan.us/v1, model name as in the table:
curl https://apiclan.us/v1/chat/completions \
-H "Authorization: Bearer sk-your-key" \
-H "Content-Type: application/json" \
-d '{"model":"deepseek-v4-pro","stream":true,"messages":[{"role":"user","content":"hi"}]}'
With the OpenAI Python SDK:
from openai import OpenAI
client = OpenAI(base_url="https://apiclan.us/v1", api_key="sk-your-key", timeout=120)
r = client.chat.completions.create(
model="deepseek-v4-pro",
messages=[{"role": "user", "content": "hi"}],
)
print(r.choices[0].message.content)
Create the API key under the DeepSeek group in the console. A key from another group returns model_not_found for these models. For long generations, use https://api.apiclan.us/v1, which avoids the 100-second proxy limit.
The same balance pays for Claude, GPT, Gemini and DeepSeek. Top-ups are in USDT, at 1 USDT for 2 credits of balance, with no monthly fee.
No subscription, no monthly minimum, no sales call. Top up with USDT and spend what you use — 1 USDT gives you 2 credits of API balance.
Read the 30-second quickstartLanguage: English · 简体中文
Prices on this page are filled in automatically from live billing data. Third-party terms are quoted from that party's own documentation.