DeepSeek

DeepSeek models

Version comparison

Last verified Oct 17, 2026. Figures that could not be confirmed on an official vendor page are marked unverified — we do not guess.

Full DeepSeek comparison

Which one should I use?

API: use deepseek-flash (DeepSeek-V4.1-Flash-1017) for most work. It is cheaper, has vision, and is the name DeepSeek currently documents as the Flash model.
Use deepseek-v4-pro when you want the Pro checkpoint. DeepSeek says it will keep serving V4 Pro after 14 Sep 2026 with the same billing until further notice.
Consumer chat-app daily limits are unverified on the API docs. Hosted DeepSeek processes data in China — see the catalog privacy note.

DeepSeek V4.1 Flash

Current

Context: 1M
Max output: 384K
API input / 1M: $0.15 / 1M
API output / 1M: $0.6 / 1M

Best for: Budget API use and anyone comparing open-weight / China-hosted options, after reading the privacy note.

API is pay-as-you-go. Free consumer chat-app limits were not listed on the API pricing page.

Source: Models & pricing

DeepSeek V4 Pro

Current

Context: 1M
Max output: 384K
API input / 1M: $0.66 / 1M
API output / 1M: $1.98 / 1M

Best for: API jobs where you specifically want the Pro checkpoint.

API is pay-as-you-go. Consumer chat-app mapping was not listed on the API pricing page.

Source: Models & pricing

Version	Context	Max out	API in / 1M	API out / 1M	Source
DeepSeek V4.1 Flash Current`deepseek-flash`	1M	384K	$0.15 / 1M	$0.6 / 1M	Models & pricing
DeepSeek V4 Pro Current`deepseek-v4-pro`	1M	384K	$0.66 / 1M	$1.98 / 1M	Models & pricing

DeepSeek V4.1 Flash

Budget API use and anyone comparing open-weight / China-hosted options, after reading the privacy note.

Very cheap API at off-peak cache-miss rates.
1M context and 384k max output.
Vision is supported on Flash; Pro’s table says vision is not.

Caveats

Peak hours cost 2×.
Hosted service data-handling is a catalog disclosure item (China).
Consumer chat limits unverified.

Off-peak cache-miss. Peak cache-miss is $0.30. Off-peak cache-hit is $0.003. Peak hours: 01:00–04:00 and 06:00–10:00 UTC, weekdays excluding Chinese public holidays.

DeepSeek V4 Pro

API jobs where you specifically want the Pro checkpoint.

Same 1M context and 384k max output as Flash.
Still officially served after 14 Sep 2026, per DeepSeek’s pricing notes.

Caveats

No vision on Pro according to the pricing table.
Peak hours cost 2×.

Off-peak cache-miss. Peak cache-miss is $1.32. Off-peak cache-hit is $0.022.

Catalog pages

DeepSeek