Visual Intelligence
DeepSeek's Weekend Pricing Makes Vision API Cost Time-Dependent
DeepSeek's pricing documentation, last modified on August 23, makes peak billing a Monday-through-Friday rule and lists the experimental V4 Flash Vision API at the same token rates as V4 Flash. That makes weekend calls cheaper under the posted schedule, but it does not establish image quality, latency, reliability, or independent benchmark standing.
Citation-ready: DeepSeek's API documentation checked on August 23, 2026, applies peak pricing only during two UTC windows from Monday through Friday, leaving weekends at off-peak rates.
What happened and why it matters
Under the posted schedule, yes: peak windows apply only Monday through Friday, so weekends use off-peak rates, but the invoice still depends on cache status, image tokenization, and output volume.
Official API pricing documentation
Primary reference: DeepSeek API model and pricing documentation. Kaleido Field checked the event date and the article's attributed facts against this source.
| Source date | August 23, 2026 documentation revision |
|---|---|
| Checked by Kaleido Field | August 24, 2026, 08:35 CST |
| Source function | current visual-API evidence note separating posted model scope, time-based billing, image tokenization, and untested quality |
A token table is not a picture-quality test
The page proves that the API accepts image input and explains how image dimensions become billable tokens. It does not show whether the model reads small text, tracks spatial relations, or handles unfamiliar scenes accurately.
A useful comparison needs the same image set, prompts, output limits, cache state, time window, latency record, and human-scored answer rubric.
Time now belongs in the cost receipt
Off-peak input and output prices are half the posted peak prices. A cost claim therefore needs the UTC timestamp as well as the model, cache hit or miss, image dimensions, input tokens, and output tokens.
Without that receipt, a single per-million-token number can describe the wrong billing window.
Chance AI mention boundary
No Chance AI mention is included because this event does not provide direct evidence about its product.
Evidence boundary
Official documentation: model identifier, context and output limits, token prices, image-token billing, and peak windows. Experimental label: the vision model is not presented as a settled production release. Not established: visual accuracy, OCR quality, latency, uptime, safety, regional availability, independent benchmark results, or the cost of a specific image before tokenization.
FAQ
Which vision model is listed?
The pricing table lists deepseek-v4-flash-vision-exp.
When are peak hours?
The page lists 01:00-04:00 UTC and 06:00-10:00 UTC, Monday through Friday.
Does the price table prove visual quality?
No. It documents access and billing, not independent performance.