DeepSeek

DeepSeek V4 Pro 0813 API pricing.

DeepSeek V4 Pro 0813 is DeepSeek’s model on the API market, released 12 Aug 2026. Prices are per million tokens in USD, as last read from the price list.

$0.12Input / 1M
$4.00Output / 1M
1.0MContext
944KMax output
Capability scores
Epoch AI
Maths3 of 4 benchmarks63
Science and knowledge2 of 2 benchmarks71
CodingNot tested by Epoch
Reasoning and puzzles2 of 2 benchmarks40
Agents and long tasksNot tested by Epoch

Each bar is this model's score in that area. The white tick marks the best score any model Epoch has tested reaches there.

Epoch Capabilities Index
156
Range 154 to 158 · #35 of 271 models Epoch has scored
Maths
63average of 3 of 4 Epoch results
FrontierMath (tiers 1 to 3) (max)65%
FrontierMath (tier 4) (max)27%
OTIS Mock AIME (max)99%
Science and knowledge
71average of 2 of 2 Epoch results
GPQA Diamond (max)92%
SimpleQA Verified (max)53%
Reasoning and puzzles
40average of 2 of 2 Epoch results
Chess puzzles (max)46%
Mystery game puzzles (max)43%

The index combines many benchmarks into one general-capability score; the range is how sure Epoch is of the number. Each area score is this site's average of the model's Epoch results in that area, where 0 is what guessing would score and 100 is perfect. Compare every model, and value for money →

Source: Epoch AI, Capabilities & benchmarking (licensed CC BY 4.0). Data refreshed 2 November 2026. These are Epoch's own tests; other benchmark sites can rank models differently.

Cheaper models that come close
Our working on Epoch's index
  1. GLM-5.3-Flash · Z.ai (Zhipu AI)index 153.22 (3 below) · 96% cheaper
  2. DeepSeek V4 Flash 0731 · DeepSeekindex 154.94 (1 below) · 86% cheaper
  3. DeepSeek V4.1 Flash · DeepSeekindex 156.11 · 85% cheaper
  4. Claude Haiku 5.5 · Anthropicindex 160.37 · 82% cheaper

Models within 5 points of DeepSeek V4 Pro 0813 on Epoch's Capabilities Index that cost less, comparing a typical chat mix (three parts input to one part output). A few points of index is a small difference and says nothing about your task: a place to start testing, not a verdict. Capability against price for every model.

Specification
USD per million tokens
Provider
DeepSeek
Model id
deepseek/deepseek-v4-pro-0813
Input
$0.12
Output
$4.00
Cached input
$0.11
Context window
1.0M tokens
Maximum output
944K tokens
Takes
text
Returns
text
Tool use
Yes
Reasoning
Yes
Released
12 Aug 2026
First tracked
17 Sept 2026

At $0.12 per million input tokens, DeepSeek V4 Pro 0813 is dearer than 75 of the 340 models on the price list.

Price changes
Tracked since 17 Sept 2026
DeepSeek V4 Pro 0813 price per million tokens over timeFree$2.50$5.0017 Sept2 Nov
InputOutput
  1. Input $0.66 → $0.12 · Output $1.98 → $4.00

Want an email when this model's price changes? Tell us →

More from DeepSeek
Cheapest input first
  1. DeepSeek V4.1 Flash$0.045 in · $0.50 out
  2. DeepSeek V4 Flash 0423$0.091 in · $0.18 out
  3. DeepSeek V4 Flash 0731$0.12 in · $0.24 out
  4. DeepSeek V4.1 Flash 1017$0.15 in · $0.60 out
  5. DeepSeek V4 Flash Vision Exp$0.22 in · $0.65 out
  6. DeepSeek V3.1$0.25 in · $0.95 out
  7. DeepSeek V3$0.26 in · $1.03 out
  8. DeepSeek V3.2$0.26 in · $0.38 out

Prices and specifications come from OpenRouter’s catalogue, last read 2 Nov 2026. A provider’s own price may differ; how this is collected.