Last verified
COST DISRUPTORFOUNDED 2023HANGZHOU · CHINAMIT OPEN WEIGHTSHIGH-FLYER BACKED

DeepSeek API Pricing

The Chinese lab behind the DeepSeek open-weight models. Spun out of quant hedge fund High-Flyer in 2023 by Liang Wenfeng, it shocked the industry by matching frontier reasoning quality at a fraction of the cost — and shipping the weights under an MIT license.

Production models
3
2 GA · 1 preview
Founded
2023
by High-Flyer fund
Cheapest tier
$0.44/M
DeepSeek V4 Flash input
Cost vs frontier
~11×
cheaper input vs GPT-5.5 at peak
Context window
1M
DeepSeek V4 Flash
Open weights
MIT
on Hugging Face
§ 01 / LINEUP

The full roster.

Side-by-side →
§ 02 / SHELF

All side-by-side.

Methodology →
Model Input /M Output /M Cached Context Max output Vision Tools Tier
DeepSeek V4 Flash $0.44 $1.32 $0.014−97% 1M Active
DeepSeek V4 Pro FLAGSHIP $1.32 $3.96 $0.044−97% 1M Frontier
DeepSeek V4 Flash Vision (Exp) PREVIEW $0.44 $1.32 $0.014−97% 1M Preview
DeepSeek V3 RETIRED Retired Apr 26, 2026 → deepseek-v4-flash Retired
DeepSeek R1 RETIRED Retired Apr 26, 2026 → deepseek-v4-pro Retired
§ 03 / PRICE CURVE

Pricing across the lineup.

How DeepSeek priced 4 models · DEC 24 → APR 26.

oldest → newest →
$5$3.75$2.5$1.25 $1.1 $0.27 $2.19 $0.55 $1.32 $0.44 $3.96 $1.32 DeepSeek V3 DEC 24 DeepSeek R1 JAN 25 DeepSeek V4 Fla… APR 26 DeepSeek V4 Pro APR 26
Input · newest $1.32/M
Output · newest $3.96/M
Each point is a model at its listed $/M price.

The DeepSeek story so far.

A short, loud history: V3 → R1 → V4, with V4 arriving 484 days after V3. Sourced from DeepSeek's API docs, model cards, and contemporaneous coverage.

JUL 31 · 2026
V4 Flash checkpoint swapdeepseek-v4-flash begins serving DeepSeek-V4-Flash-0731; same API id, same prices
RELEASE
JUL 24 · 2026
Legacy ids retireddeepseek-chat and deepseek-reasoner become inaccessible at 15:59 UTC; V4 Flash and V4 Pro are the only chat ids left
RETIRED
MAY 31 · 2026
V4 Pro promo expiry passes without a reprice — the announced May 31 revert to $1.74/$3.48 never took effect, and the flat $0.435/$0.87 card stayed live for another eleven weeks
PRICING
APR 26 · 2026
DeepSeek V4 Pro + V4 Flash launched (preview) — MIT open weights, 1.6T-param MoE, 1M context
RELEASE
APR 26 · 2026
Pricing reset — cache-hit input cut to 1/10 of launch; legacy deepseek-chat/deepseek-reasoner fold into V4; V3 + R1 retired
PRICING
JAN 20 · 2025
DeepSeek R1 released — o1-class reasoning at a fraction of the price; the consumer app tops the US App Store and rattles AI markets
RELEASE
DEC 26 · 2024
DeepSeek V3 released — 671B-param MoE open weights trained at a breakthrough-low compute cost
RELEASE
JUL · 2023
DeepSeek founded by Liang Wenfeng as an AI lab funded by quant hedge fund High-Flyer, Hangzhou
CORPORATE
§ 04 / ACCESS

Where to get it.

Methodology →
§ 04 / BEST FOR

Which DeepSeek for what.

More scenarios →
If you need frontier reasoning at minimal cost
DeepSeek V4 Pro
Profile →
If you want a cheap high-volume daily driver
DeepSeek V4 Flash
Profile →
If you process 1M-token long documents
DeepSeek V4 Flash
Profile →
If you must self-host for data control
V4 weights (MIT)
Hugging Face →
If you need non-China data residency
V4 on a Western host
Hosts →
If you're cost-cutting from GPT-5 / Claude
DeepSeek V4 Flash
Profile →
§ 06 / BACKGROUND

The company behind it.

www.deepseek.com →

DeepSeek was founded in July 2023 in Hangzhou by Liang Wenfeng, who also co-founded and runs the quantitative hedge fund High-Flyer (started in 2015–2016). DeepSeek is owned and funded by High-Flyer rather than by external venture capital — an unusual structure that let it train large models on a GPU cluster the fund had already built for trading.

The lab's signature is compute efficiency. Its mixture-of-experts (MoE) training recipes delivered frontier-class quality at a small fraction of the budgets reported by US labs. Liang has reportedly held a controlling personal stake (~84% as of 2024), and the team is famously lean — on the order of ~150 people, with many hired straight out of university.

DeepSeek's breakout moment came with V3 (December 2024) and then R1 (January 2025): R1 matched OpenAI o1-class reasoning at roughly 1/27 the price, the consumer app briefly topped the US App Store, and the release triggered a sharp sell-off in AI hardware stocks. In April 2026 the lab shipped the V4 family — a 1.6T-parameter MoE (V4 Pro) and a 284B MoE (V4 Flash), both with 1M-token context and MIT-licensed open weights.

Pricing is the headline, and since August 16, 2026 it depends on the hour: V4 Pro runs at $1.32/$3.96 per M at peak and $0.66/$1.98 off-peak. The old flat $0.435/$0.87 was a 75% launch promo that was supposed to revert to $1.74/$3.48 on May 31, 2026 but never did — and V4 Flash at $0.44/$1.32 peak, $0.22/$0.66 off-peak — still well below frontier US models, with cache-hit input at about 1/31 of a miss.

The trade-offs are real: the lab was text-only until August 2026 and is barely past it — the one model that reads images, V4 Flash Vision, ships under an experimental id, and nothing here handles audio. The company is China-based with data stored in China by default, and US export-control and procurement-policy questions apply. The MIT weights are the escape hatch — teams that need US/EU residency can self-host. Versus OpenAI and Anthropic, DeepSeek trades multimodal breadth and Western data governance for radically lower cost and full open weights.

§ 07 / COMPETITORS

Other frontier labs.

All providers →

Frequently asked.

Practical questions about DeepSeek pricing, open weights, and data residency.

Q · 01 Which DeepSeek model should I start with? +
For most workloads: DeepSeek V4 Flash ($0.44/$1.32 peak, $0.22/$0.66 off-peak) — it covers both non-thinking and thinking modes cheaply. Step up to V4 Pro ($1.32/$3.96 peak, $0.66/$1.98 off-peak) for the hardest reasoning. Both ship a 1M-token context. See the use case picker above.
Q · 02 How much cheaper is DeepSeek than OpenAI or Anthropic? +
On input, V4 Pro is roughly 3.8× cheaper than GPT-5.5 at peak ($1.32 vs $5) and ~7.6× cheaper on output ($3.96 vs $30). Off-peak, which is 17 hours of every weekday and all weekend, both gaps double. With cache-hit input at $0.044/M the gap widens further still. The honest caveat used to be that DeepSeek is text-only; since August 2026 there is one exception, the experimental V4 Flash Vision, and it is still narrower than what Google or OpenAI accept.
Q · 03 Is there an off-peak discount? +
Yes — and it is the whole billing model now, not a discount bolted on top. Since 16:00 UTC on August 16, 2026 DeepSeek charges a peak rate during 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday, and exactly half that in every other hour, weekends included. So off-peak covers 17 hours of a weekday and all of Saturday and Sunday. We store the peak figures because DeepSeek describes off-peak as "half the peak rates" — the peak card is the list price. Anything that can be scheduled overnight or at the weekend costs half. Re-verified on the vendor page August 23, 2026.
Q · 04 Are DeepSeek models open-weight? +
Yes. V4 Pro and V4 Flash weights are MIT-licensed on Hugging Face — fully self-hostable, commercially usable, with no per-token fee. This is what lets teams run DeepSeek outside China for data-residency reasons.
Q · 05 Where is my data stored, and is that a problem? +
Using the direct DeepSeek API, data is processed and stored in China by default, which is a blocker for many EU/US compliance and procurement teams. The workaround is to self-host the MIT weights on your own US/EU cloud, or use a Western third-party host. US export-control and policy questions also apply.
Q · 06 Did the 75% V4 Pro promo ever end? +
No. DeepSeek announced the promo would expire on May 31, 2026, reverting V4 Pro to $1.74/M input and $3.48/M output. That date passed without a reprice: we have re-verified the pricing page repeatedly since. On August 16, 2026 DeepSeek replaced the flat card with peak/off-peak billing at $1.32/$3.96 and $0.66/$1.98. We treat it as the standard rate rather than a promo, and we track the page daily in case it moves.
Q · 07 How does DeepSeek compare to Qwen and Mistral? +
Qwen (Alibaba) is the closest Chinese peer with broader multimodal and multilingual coverage; Mistral is the EU open-weight peer with GDPR-native residency. DeepSeek's edge is the lowest cost-per-token at frontier reasoning quality, with the same China-data caveat as Qwen.