DeepSeek V4 Pro 0813

Capability & preference

General capability · ECI

A cross-benchmark capability estimate. Index points, not a percentage; intervals show uncertainty.

DeepSeek V4 Pro 0813 155.9^* DeepSeek V4 Pro 0813 · 95% interval 154.2–158.1

Model performance

Radar task score: four tests, equal weight. An experimental summary of these tasks, not a universal intelligence rating.

DeepSeek V4 Pro 0813

Tested configuration Radar task score

63.0^* / 100

4/4 core tests · experimental v1

7 benchmark results for this configuration

Scientific reasoning

GPQA Diamond · 25%

Expert-written science questions; strict multiple-choice scoring. Chance accuracy is 25%.

DeepSeek V4 Pro 0813 91.7^*% Approx. 95% interval: 88.7–94.7% · 2026-08-18

Novel problem solving

Mystery Game Puzzles · 25%

Best-next-move puzzles in an undisclosed game variant. A narrow reasoning test, not general intelligence.

DeepSeek V4 Pro 0813 43.0^*% Approx. 95% interval: 33.2–52.8% · 2026-08-19

Advanced mathematics

FrontierMath Tiers 1-3 · 25%

Private v2 mathematics problems with Python tools. Earlier versions and Tier 4 are excluded.

DeepSeek V4 Pro 0813 64.6^*% Approx. 95% interval: 59.0–70.1% · 2026-08-19

Factual answers

SimpleQA Verified · 25%

Correct factual answers under a guessing prompt, graded by a model. Not a hallucination or abstention score.

DeepSeek V4 Pro 0813 52.9^*% Approx. 95% interval: 49.8–56.0% · 2026-08-27

End-to-end software

MirrorCode · Separate result

Whole-program reimplementation with hidden tests; ML + Private, two languages. Budget: seven days and ten billion tokens per attempt.

DeepSeek V4 Pro 0813 33.3^*% Approx. 95% interval: 13.7–53.0% · 2026-10-20

Competition mathematics

OTIS Mock AIME 2024-2025 · Separate result

Competition problems with model-extracted answers checked against an exact key. Earlier grading procedures are excluded.

DeepSeek V4 Pro 0813 98.6^*% Approx. 95% interval: 96.7–100.0% · 2026-08-18

Research mathematics

FrontierMath Tier 4 v2 · Separate result

The hardest private FrontierMath problems, corrected v2 set, with Python tools. Separate from Tiers 1-3 in the core score.

DeepSeek V4 Pro 0813 26.8^*% Approx. 95% interval: 13.1–40.6% · 2026-08-19

Bars show reported task scores. Whiskers are approximate 95% intervals where standard errors are available. Different reasoning budgets are not equal-cost tests. Missing results are never zero.

Run it locally

Inference · memory estimate

DeepSeek V4 Pro 0813

Not enough verified architecture information for a reliable VRAM estimate.

API cost

USD / month

Input tokens / monthOutput tokens / month

Standard text-token estimate. Excludes taxes, caching, batch discounts, reasoning and multimodal surcharges. Provider limits and regional availability may differ.

DeepSeek V4 Pro 0813

Provider offer

DeepSeek V4 Pro 0813 / Pricing tiers 5

Provider / scope	Rate	Currency	Amount / unit
DeepInfradeepseek-ai/DeepSeek-V4-Pro-0813Standard	Cached input	USD	0.1 / 1M tokens
	Input	USD	1.3 / 1M tokens
	Output	USD	2.6 / 1M tokens
DeepSeekdeepseek-v4-proOff Peak off_peak	Cached input	CNY	0.15 / 1M tokens
	Input	CNY	4.5 / 1M tokens
	Output	CNY	13.5 / 1M tokens
DeepSeekdeepseek-v4-proStandard	Cached input	CNY	0.3 / 1M tokens
	Input	CNY	9 / 1M tokens
	Output	CNY	27 / 1M tokens
DeepSeekdeepseek-v4-proOff Peak off_peak	Cached input	USD	0.022 / 1M tokens
	Input	USD	0.66 / 1M tokens
	Output	USD	1.98 / 1M tokens
DeepSeekdeepseek-v4-proStandard	Cached input	USD	0.044 / 1M tokens
	Input	USD	1.32 / 1M tokens
	Output	USD	3.96 / 1M tokens

Capabilities

Context window: 1,000,000 tokens
Maximum output: 384,000 tokens
Reasoning: Yes
Tool calling: Yes
Open weights: Yes
Model license: MIT

Input modalities

text

Output modalities

text

Processing & platform

Architecture: deepseek_v4