DeepSeek V4 API pricing on OpenAny: compare routes, peak hours and total cost

DeepSeek can make a model-heavy agent economical, but a single price such as "$0.675" is not enough to budget the task. You need the model variant, input/output split, selected route and time window.

OpenAny is useful because you can compare provider routes and combine eligible model usage with external API spending. The recommendation should rest on that visibility and workflow fit, not an unverified promise of all-day off-peak pricing.

Start with the exact model

DeepSeek's official page distinguishes deepseek-flash from deepseek-v4-pro. It also explains that certain legacy Flash IDs now resolve to a newer Flash version. A route with "V4" in its name is therefore not automatically the same model or snapshot as another route.

The table below compares V4-Pro only, in USD per one million tokens, using standard uncached input and output. OpenAny values use the public customer fields, not partner-pricing fields. Checked November 4, 2026.

Access path	Base/off-peak input / output	Peak input / output
DeepSeek official V4-Pro	$0.66 / $1.98	$1.32 / $3.96
OpenAny V4-Pro, Tencent route	$0.620706 / $1.862119	$1.241413 / $3.724238
OpenAny V4-Pro, DeepSeek route	$0.651724 / $1.955172	$1.303448 / $3.910345

The first OpenAny route's base rates are about 5.95% below the official off-peak reference. The second is about 1.25% below. These are rate comparisons, not guaranteed workload savings or proofs of identical upstream service conditions.

Check the clock and the calendar

DeepSeek's official peak schedule is Monday through Friday, 01:00-04:00 and 06:00-10:00 UTC, with Chinese public holidays excluded. Other hours are off-peak.

The reviewed OpenAny V4-Pro timePricingRules also specify weekday UTC windows, and publish higher prompt and completion values for those windows. The inspected rules do not explicitly encode the official holiday exclusion. Do not assume matching holiday treatment without an execution-time quote or provider confirmation.

Other OpenAny DeepSeek/Flash routes publish different schedules. Never apply one provider's time rule to the whole family. A catalog's base value is not an all-day price commitment.

Estimate both scenarios

Illustrative workload: 10 million uncached input tokens and 2 million output tokens, all in one price window. No caching, tools, fees or retries are included.

Scenario	Official V4-Pro	OpenAny Tencent route
Off-peak/base	$10.56	$9.93
Peak	$21.12	$19.86

OpenAny base: 10 x 0.620706 + 2 x 1.862119 = $9.931298
OpenAny peak: 10 x 1.241413 + 2 x 3.724238 = $19.862606

For mixed traffic, calculate each time window separately and add the results. For cache-heavy traffic, use the cache-read rate only for tokens actually billed as cache hits. A claimed input rate says nothing about the cost of generating a long answer.

What happened to the $0.675 claim?

It is not used as the article's current price. The public snapshot does not establish that exact V4-Pro customer rate as a flat, all-day offer. The two reviewed routes have explicit peak rules. Publishing the number without those qualifications would be misleading, even if it appeared in an earlier catalog or proposal.

Before a production task, inspect OpenAny's current model routes, select the intended provider where supported, and approve the complete task budget. If the client cannot express provider selection, establish the default route rather than pricing the task against a route it may not use.

Use with your Agent

Refresh prices and calculate your token mix before authorizing a paid test.

Read https://openany.si/AGENT.md and follow its setup and credential-safety instructions.

Task: Compare DeepSeek V4-Pro pricing for 10 million uncached input and 2 million output tokens. Validate quantities and refresh DeepSeek's official price page plus OpenAny's public model prices. Record exact model, provider route, version where available, USD per-million rates, UTC time windows, weekday/holiday handling and check date. Use prompt and completion customer fields and matching timePricingRules; never substitute partner prices, cache-hit prices or a Flash model for V4-Pro. Calculate peak and off-peak scenarios separately and flag differences in schedule coverage. Do not claim a flat all-day $0.675 rate unless a current exact route proves it. Return a dated comparison, formulas and a bounded OpenAny recommendation. Make no authenticated request, payment, top-up or configuration change.

Guide: https://openany.si/guides/deepseek-v4-pricing-openany/
Ask for a budget and approval before billable requests. Never request credentials in chat. Do not substitute providers or execute other actions without approval. Preserve sources and timestamps; report missing data and actual results honestly.

To compare other models, read Qwen overseas pricing and Claude API pricing. For access and funding, create an account and review USDC top-ups.

Sources

• DeepSeek official models, versions and pricing
• OpenAny customer route prices and time rules
• OpenAny model execution guidance