DeepSeek cut V4.1 Flash API prices on September 10, 2026
Changed: Sep 10, 2026
Published
At 04:00 UTC on September 10, 2026, DeepSeek lowered Flash with the V4.1 release. Peak cache-miss input fell from $0.44 to $0.30 and output from $1.32 to $1.20 per 1M tokens. Pro rates didn't change.
| What changed | Before | Now |
|---|---|---|
| Flash input (cache miss), peak, per 1M tokens | $0.44 | $0.30 |
| Flash output, peak, per 1M tokens | $1.32 | $1.20 |
| Flash input (cache hit), peak, per 1M tokens | $0.014 | $0.006 |
| Flash input (cache miss), off-peak, per 1M tokens | $0.22 | $0.15 |
| Flash output, off-peak, per 1M tokens | $0.66 | $0.60 |
| Flash input (cache hit), off-peak, per 1M tokens | $0.007 | $0.003 |
| Pro input (cache miss), peak, per 1M tokens | $1.32 | Unchanged |
| Pro output, peak, per 1M tokens | $3.96 | Unchanged |
What changed
The pricing page archived on September 9, 2026 lists Flash at $0.44 per 1M cache-miss input tokens and $1.32 for output at peak. Off-peak it lists $0.22 and $0.66. The copy archived on September 11 lists $0.30 and $1.20 at peak, and $0.15 and $0.60 off-peak. Cache-hit input fell from $0.014 to $0.006 at peak.
DeepSeek’s release post for V4.1-Flash says the new pricing takes effect at 04:00 UTC on September 10, 2026. The live page, checked November 17, still shows these Flash rates under the model name deepseek-flash.
The cut is 32% on cache-miss input and 9% on output. Pro rates stayed where the August 16 repricing left them. The release post adds that from 04:00 UTC on September 14, 2026, requests to deepseek-v4-pro route to V4.1-Flash at Flash rates until V4.1-Pro launches, so check which rate your Pro calls are billed at.
Who feels it
Input-heavy workloads gain most: long prompts, retrieval, document reading. Output-heavy chat saves little, since output dropped only 9%.
Flash is still pricier than the flat rates that ended on August 16. Peak input of $0.30 compares with the old $0.14.
What to do
- Update any Flash estimate that uses $0.44 and $1.32.
- Use the
deepseek-flashmodel name. DeepSeek listsdeepseek-v4-flashas a legacy name that is billed at the Flash price. - Keep batch work in the off-peak window, which is half the peak rate.
Compare current API prices: check current API token costs for GPT, Claude, Gemini, Grok, DeepSeek and Mistral.
Sources
- DeepSeek API pricing page, archived September 9, 2026 (Wayback Machine)
- DeepSeek API pricing page, archived September 11, 2026 (Wayback Machine)
- DeepSeek announcement: DeepSeek-V4.1-Flash release, new pricing from 04:00 UTC Sept 10, 2026
- DeepSeek API pricing page, live
How we check prices: our method. Spotted a price we got wrong? Tell us at the contact page and we correct it in a dated note like this one.
Pricing notes
Get the monthly roundup of price changes
One short email a month with every price change we confirmed. No promos in between.