DeepSeek V4 Pro API pricing per million tokens
DeepSeek V4 Pro costs $1.32 per million input tokens and $3.96 per million output, with a 1M context window.
AIGROW credit covers Claude Opus 5.5, Claude Fable 5.2, Claude Fable 5.1, GPT-6 Astra, Claude Sonnet 5.5, GPT-6.1 Sol, GPT-6.1 Terra, Grok 4.7 and Gemini 4 Argon. See API credit
$25 a month brings $110 of API credit and cancels any time; $25 once buys $100. US dollars per million tokens, standard rate, shortest context band. Number 69 of 256 on the Epoch Capabilities Index by Epoch AI, data of 12 November 2026. CC BY 4.0
- 1,000 support replies
- $5.54
- Cached input
- $0.044
- Context window
- 1,000,000 tokens
- Released
- 13 Aug 2026
- Weights
- Open weights
From DeepSeek’s own pages, 12 November 2026.
DeepSeek V4 Pro costs $1.32 per million input tokens and $3.96 per million output tokens on DeepSeek’s own API, and $0.044 for cached input. At those prices 1,000 support replies cost $5.54, the 24th most expensive of 49 priced models on our list.
AIGROW API credit per million input / output tokens: Claude Opus 5.5 $1.60 / $8.00, Claude Fable 5.2 $3.20 / $16.00, Claude Fable 5.1 $4.00 / $20.00, GPT-6 Astra $2.50 / $12.50, Claude Sonnet 5.5 $0.80 / $4.00, GPT-6.1 Sol $0.50 / $2.50, GPT-6.1 Terra $0.25 / $1.25, Grok 4.7 $0.50 / $1.50 and Gemini 4 Argon $1.00 / $5.00. That is up to 75% below list price. See API credit
What a job costs
Three jobs, 1,000 runs of each, at the list price above, and where each puts DeepSeek V4 Pro among the 49 models with a list price.
| Job | 1,000 runs | Among 49 |
|---|---|---|
| Support reply3,000 in, 400 out. | $5.54 | 24th most expensive |
| Document summary25,000 in, 1,000 out. | $36.96 | 24th most expensive |
| Agent task80,000 in (70,000 cached), 3,000 out. | $28.16 | 19th cheapest |
The token counts are ours, chosen to look like real work; the prices are DeepSeek’s. All three jobs stay under 100,000 tokens a request, below every long-context surcharge on the list.
DeepSeek V4 Pro on real small-business tasks
What the AIGROW benchmark measured when DeepSeek V4 Pro was given 20 office tasks, each marked pass or fail by fixed checks.
- Success rate
- 90% of 20 runs passed (95% interval 70 to 97%), 4th of 36 models.
- Cost per 1,000 passes
- $5.14, failed runs included.
- Answer time
- 21 s median, 49 s at the 90th percentile.
- Strongest and weakest
- Appointment scheduling (100% passed) and quotes (50%).
The small print
What the base rate above leaves out, copied from the maker’s pages: surcharges, discounts, promotions and cache pricing.
Peak-hour rate, recorded as the list price. Off-peak (all hours outside 01:00-04:00 and 06:00-10:00 UTC on weekdays) is half: $0.66 / $1.98, cache hit $0.022. Peak and off-peak pricing started 2026-08-16. API name deepseek-v4-pro serves DeepSeek-V4-Pro-0813 (V4 first shipped 2026-04-24). Max output 384K.
DeepSeek V4 Pro against the models it is weighed with
Each comparison puts both list prices side by side and costs the same three jobs on each.
- vs Mistral Large 4
- Frontier APIs from Europe and China, with Mistral at about half DeepSeek's peak rate and close to its off-peak one.
- vs Claude Opus 5.5
- A low-cost open-weight frontier model against Anthropic's flagship, at about a third of the input price.
- vs Claude Opus 5
- A low-cost open-weight frontier model against a premium closed model many businesses still run.
- vs GPT-5.6 Sol
- Shows how much quality a buyer gives up, if any, for a model that costs a fraction of Sol.
- vs Qwen3.8 Max
- Two Chinese frontier APIs that buyers weigh against each other, both far cheaper than the US flagships.
- vs GLM-5.3
- Open-weight Chinese frontier models in the same price range.
- vs MiniMax M3
- Two of the cheapest models that their makers position as frontier, both open weight.
Other DeepSeek models
| Model | Input | Output | Context | 1,000 replies |
|---|---|---|---|---|
| DeepSeek V4.1 FlashDeepSeek · open weights | $0.30 | $1.20 | 1M | $1.38 |
Where the price comes from
Read on 12 November 2026. Each price is the standard rate for the shortest context band, with no batch or priority discount; long-context surcharges, batch rates and promotions are in each model’s small print. No price was taken from a reseller.
Compare DeepSeek V4 Pro with 53 other models
The price list sets every model’s input and output price side by side, with what a support reply, a document summary and an agent task cost on each.
Prices checked 12 November 2026 · See them by release date