DeepSeek V4.1 Flash-1017: Same Price, Half Off-Peak
Short answer. On October 17, 2026 DeepSeek upgraded V4.1 Flash to DeepSeek-V4.1-Flash-1017. The API name (deepseek-flash) and the price did not change: $0.30 input and $1.20 output per million tokens at peak, half that off-peak. Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday, excluding Chinese public holidays.
In plain words. If you use DeepSeek’s app or website, DeepSeek says they now run the new version too; this site does not track app prices. If you pay for the API, the same work costs half outside peak hours, including all weekend, so long batch jobs are cheaper then.
On October 17, 2026 DeepSeek replaced the model behind its deepseek-flash
API name with a new version, DeepSeek-V4.1-Flash-1017. The name you call
did not change and neither did the price. This post sets out what
DeepSeek’s own pages say about the update, both read on October 19, 2026:
the price table with its peak and off-peak rates, what one job costs in
each window, a Search and Fetch billing date that matters to Claude Code
and Codex users, and the old model names that now point at the new
version.
What changed on October 17
DeepSeek’s change log, under “Date: 2026-10-17”, says: “DeepSeek V4.1 Flash has been upgraded to DeepSeek-V4.1-Flash-1017. The official web and APP models have also been updated to the new version. The API calling method and pricing remain unchanged — simply use deepseek-flash to access the latest model.”
The pricing page agrees. Its model table now reads “MODEL VERSION | DeepSeek-V4.1-Flash-1017 | DeepSeek-V4-Pro-0813 |”; on October 1, 2026 the same line read “DeepSeek-V4.1-Flash” with no suffix. V4 Pro keeps its 0813 version. Both columns print the same figures as before the update, which is why this site’s price-change log has no entry for it: no price moved.
The price table, peak and off-peak
DeepSeek prints one column per model and two rates per price, off-peak and peak. As the page printed them on October 19, 2026, per million tokens:
| Per 1M tokens | V4.1 Flash, off-peak | V4.1 Flash, peak | V4 Pro, off-peak | V4 Pro, peak |
|---|---|---|---|---|
| Input, cache hit | $0.003 | $0.006 | $0.022 | $0.044 |
| Input, cache miss | $0.15 | $0.30 | $0.66 | $1.32 |
| Output | $0.60 | $1.20 | $1.98 | $3.96 |
DeepSeek’s page prints the same figures without trailing zeros ($0.3, $1.2).
The page writes the rows this way: “(CACHE MISS) | OFF-PEAK | $0.15 | $0.66 |” followed by “PEAK | $0.3 | $1.32 |”, and “1M OUTPUT TOKENS | OFF-PEAK | $0.6 | $1.98 |” followed by “PEAK | $1.2 | $3.96 |”. Its second footnote sets the hours: “Off-peak rates are half of the peak rates. Peak hours are 01:00 - 04:00 and 06:00 - 10:00 UTC, Monday through Friday, excluding Chinese public holidays. All other hours are off-peak, including weekends and Chinese public holidays in full.”
That is seven peak hours on a working day, 35 of the 168 hours in a week before any holiday is taken out. This site lists the peak rate as the price on both model pages, because the off-peak rate is defined against it (“half of the peak rates”), and shows the off-peak rate next to it.
One job, two windows
Take one batch of work: 30M input tokens read from the cache, 10M input tokens that miss it, and 2M output tokens. At DeepSeek’s rates the bill is:
- V4.1 Flash, peak: 30 × $0.006 = $0.18, plus 10 × $0.3 = $3.00, plus 2 × $1.2 = $2.40. Total $5.58.
- V4.1 Flash, off-peak: 30 × $0.003 = $0.09, plus 10 × $0.15 = $1.50, plus 2 × $0.6 = $1.20. Total $2.79.
- V4 Pro, peak: 30 × $0.044 = $1.32, plus 10 × $1.32 = $13.20, plus 2 × $3.96 = $7.92. Total $22.44.
- V4 Pro, off-peak: 30 × $0.022 = $0.66, plus 10 × $0.66 = $6.60, plus 2 × $1.98 = $3.96. Total $11.22.
Every line halves, as the footnote says it should. The same work costs about four times as much on V4 Pro as on V4.1 Flash in either window, and running V4 Pro off-peak still costs about twice as much as running V4.1 Flash at peak. The page does not say how a request that starts in one window and ends in the other is billed, and this post does not guess.
Search and Fetch from Claude Code and Codex, free until November 1
The same page now prices two standalone tools, “Search | Per 1,000 successful requests | $0.75 |” and “Fetch | Per 1,000 successful requests | $0.75 |”, and adds: “Note: when you use the DeepSeek API through DeepSeek Harness, Claude Code, or Codex, server-side tool calls already hit the Search and Fetch APIs. These calls are currently free of charge; per-request billing starts at 00:00 Beijing time on November 1, 2026.”
Two things follow from the wording alone. The free period is about those Search and Fetch calls, billed per request, not about the token prices in the table above, which apply as usual. And the start of billing is given in Beijing time; this post quotes it as printed and does not convert it.
The old names still answer
DeepSeek’s first footnote covers the names it retired in September: “(1) Use deepseek-flash as the model name. The legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted, but the corresponding models have been retired, their requests are served by the DeepSeek-V4.1-Flash-1017 model and billed at the Flash price.” This site records both of those models as retired on September 10, 2026, the day V4.1 Flash was released, and lists them on Deprecations. A request sent under either old name today reaches the 1017 version and pays the V4.1 Flash rates above.
Where this is recorded
The DeepSeek V4.1 Flash page and the DeepSeek V4 Pro page carry the figures above with the date they were read and a link to DeepSeek’s page. The new version name is recorded as an alias of the V4.1 Flash row, so this site’s weekly check matches it. The price-change log will show an entry only if one of these rates moves. Peak and off-peak pricing in general, and the other things that decide a bill, are explained in how LLM API pricing works.
Frequently asked questions
Did DeepSeek change the price of V4.1 Flash with the 1017 update?
No: DeepSeek's change log entry of October 17, 2026 says "The API calling method and pricing remain unchanged", and on October 19, 2026 the pricing page still prints $0.3 input and $1.2 output per million tokens at peak.
When are DeepSeek's peak hours?
From 01:00 to 04:00 and from 06:00 to 10:00 UTC, Monday through Friday, excluding Chinese public holidays, as DeepSeek's pricing page prints them. All other hours are off-peak and cost half.
How much does the same job cost at peak and off-peak on DeepSeek?
A job of 30M cached input, 10M uncached input and 2M output tokens costs $5.58 at peak and $2.79 off-peak on V4.1 Flash, and $22.44 and $11.22 on V4 Pro.
Are Search and Fetch calls from Claude Code free on DeepSeek?
DeepSeek's pricing page says the server-side Search and Fetch calls made through DeepSeek Harness, Claude Code or Codex are currently free of charge, and that per-request billing starts at 00:00 Beijing time on November 1, 2026. The note is about those per-request APIs, not about token prices.
Drafted with AI assistance from the cited source and reviewed by Álvaro Lucero Rufino, who is responsible for what stays published here. Automated publishing was discontinued in September 2026; posts that no longer met that standard were removed rather than left up. This is original explanatory writing, not a reproduction — read the source above for the full story.