DeepSeek V4 Flash Vision API: Pricing, Off-Peak Rate and Setup

DeepSeek V4 Flash Vision is the experimental vision sibling of DeepSeek V4 Flash, live as model ID deepseek-v4-flash-vision-exp. RouterPlex bills $0.22 per 1M input tokens and $0.66 per 1M output tokens at all hours.

Update, 20 October 2026: DeepSeek has retired this preview. Its pricing page says the legacy deepseek-v4-flash-vision-exp name is still accepted, but requests are served by DeepSeek-V4.1-Flash and billed at the Flash price: $0.15 / $0.60 off-peak and double that during peak hours (01:00–04:00 and 06:00–10:00 UTC, Monday to Friday). RouterPlex serves this ID through a separate upstream and keeps a flat rate; it does not switch by the clock.

DeepSeek V4 Flash Vision API pricing #

Cached input on this catalog is $0.007 per 1M.

A $5 prepaid balance at the billed rate buys roughly 22.7 million input tokens, or 7.6 million output tokens, if a workload used only one category.

Flash vs Flash Vision vs Pro #

The two Flash IDs are now close: Flash Vision costs a little more on input and a little less on output. On input-heavy agent turns with no image, deepseek-v4-flash is still the cheaper ID.

Cost per agent turn #

One 40,000-token prompt returning 2,000 tokens, no cache:

The arithmetic for Flash Vision: 40,000 × $0.22 / 1M + 2,000 × $0.66 / 1M = $0.0088 + $0.00132 = $0.01012.

Live pages: DeepSeek V4 Flash Vision, DeepSeek V4 Flash, DeepSeek V4 Pro.

DeepSeek V4 Flash Vision specifications #

Call DeepSeek V4 Flash Vision with RouterPlex #

curl https://api.routerplex.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash-vision-exp",
    "messages": [
      {"role": "user", "content": "Describe the layout of this screenshot."}
    ]
  }'

import os
from openai import OpenAI
 
client = OpenAI(
    api_key=os.environ["ROUTERPLEX_API_KEY"],
    base_url="https://api.routerplex.com/v1",
)
 
response = client.chat.completions.create(
    model="deepseek-v4-flash-vision-exp",
    messages=[{"role": "user", "content": "Extract the table from this screenshot."}],
)
 
print(response.choices[0].message.content)

Attach images with the usual OpenAI image_url content block. Claude Code can point at the same model with ANTHROPIC_BASE_URL=https://api.routerplex.com and model deepseek-v4-flash-vision-exp. See the Claude Code setup guide.

Which DeepSeek should you actually call? #

• Text-only volume — deepseek-v4-flash at $0.18 / $0.73.
• Need vision on DeepSeek — deepseek-v4-flash-vision-exp, knowing it is experimental.
• Need a generally available vision Flash — GLM-5.3 Flash or Gemini 3.8 Flash.
• Need DeepSeek Pro quality, no images — deepseek-v4-pro.

Give an experimental ID its own hard budget. The live DeepSeek V4 Flash Vision page is the number this route will actually bill. Text-only volume models are ranked on the cheapest models ledger.