Skip to model information

DeepSeek V4 Pro

Available

The stable API name for the latest DeepSeek V4 Pro version; the provider currently identifies that version as DeepSeek-V4-Pro-0813.

Access details: DeepSeek API

Undated identifier · Identity and lifecycle apply to this exact identifier and serving surface.

deepseek-v4-pro
Source check: Sep 13, 2026

Capabilities & limits

Context window
1M
Maximum output
384K
Image input
Unsupported
Tool calling
Supported
Structured output
Supported

Documented interfaces and limits. These are not measures of model quality.

Pricing

USD / 1M tokens

See the full rate card for captured prices and billing conditions.

USD per million standard text tokens. Detailed cache, batch and other pricing conditions appear in the rate card.

Full rate card 6 rates

Amounts in USD. Check the unit and conditions on each rate.

Cache-hit input, off-peak

$0.022 per 1M tokens
  • Off-peak: all times outside 01:00-04:00 and 06:00-10:00 UTC, Monday-Friday

Cache-hit input, peak

$0.044 per 1M tokens
  • Peak: 01:00-04:00 and 06:00-10:00 UTC, Monday-Friday

Cache-miss input, off-peak

$0.66 per 1M tokens
  • Off-peak: all times outside 01:00-04:00 and 06:00-10:00 UTC, Monday-Friday

Cache-miss input, peak

$1.32 per 1M tokens
  • Peak: 01:00-04:00 and 06:00-10:00 UTC, Monday-Friday

Output, off-peak

$1.98 per 1M tokens
  • Off-peak: all times outside 01:00-04:00 and 06:00-10:00 UTC, Monday-Friday

Output, peak

$3.96 per 1M tokens
  • Peak: 01:00-04:00 and 06:00-10:00 UTC, Monday-Friday

Peak/off-peak schedule effective since 2026-08-16 16:00 UTC.

Price history
Observed Sep 13, 2026

Input Not verified per million tokens

Not verified. These sources establish the identifier, not this missing fact.

Observed Sep 13, 2026

Output Not verified per million tokens

Not verified. These sources establish the identifier, not this missing fact.

Technical details

Parameters
1.6T
Active parameters
49B
Architecture
Mixture-of-Experts with DeepSeek Sparse Attention
Input modalities
text
Output modalities
text
1 detail not verified
  • Training / knowledge cutoff

How they describe it

Provider quote not captured yet.

What they say is different

Performance rivaling the world's top closed-source models.

Verbatim provider claims · not an independent assessment

Release & lifecycle

Availability applies to this exact identifier on DeepSeek API.

First API availability
Apr 24, 2026
available
Observed Sep 13, 2026

Considering a switch?

Test quality, tool behavior, latency, and billable tokens on your own tasks. Matching API features does not imply equivalent outputs.

Compare with GPT-5.6 Sol

Curated snapshot through September 13, 2026. Coverage is partial; verify current availability and pricing with the provider.