> DeepSeek-V4-Flash is an open-weights AI model from DeepSeek, released 31 Jul 2026. Index score 165 (GPT-4 = 100), rank 33 of 229 ranked models. API price $0.017 in / $1.28 out per million tokens, 1.05M context.

[Models](https://themodelindex.org/models/) /[DeepSeek](https://themodelindex.org/labs/deepseek/)

# DeepSeek-V4-Flash

DeepSeek

Released 31 Jul 2026

Open

Date verified

[Compare](https://themodelindex.org/compare/deepseek-v4-flash-2026-07-vs-claude-opus-5/) [Run it yourself →](https://themodelindex.org/local/) [Announcement](https://api-docs.deepseek.com/updates/)

165153–176

Index score

#33 of 229 models

3/ 8

Core tests taken

3 of 6 areas · plus 12 anchor tests

−22

Versus the frontier at release

best then: Claude Opus 5

Replaced by [DeepSeek-V4.1-Flash](https://themodelindex.org/models/deepseek-v4-1-flash/)

$0.017/ $1.28

Price per million tokens

input / output, list price

1.05M

Context window

tokens it reads at once

## Where it sits · on the frontier

_Chart: Model Index score of every scored model since late 2022, with the best closed and best open model over time_

Claude Opus 4.6 reached this score 6 months earlier.

## Test results · by area · tick = best by any model

Show 12 anchor tests

Reasoning & knowledge

evidence

GPQA Diamond

91%

Humanity's Last Exam

no result

Maths

evidence

FrontierMath T1-3

58%

Coding

evidence

SWE-bench Verified

no result

Terminal-Bench

no result

Agents

evidence

APEX-Agents

no result

Novel problems

evidence

ARC-AGI-2

61%

Long tasks

no evidence yet

METR time horizon

no result

Independent results from the Epoch AI benchmarking hub, best reasoning setting. The core tests are shown here; anchor tests also inform the score and can be shown above.

## What others say · rank on each leaderboard

The Model Index

independent tests, our method

#33

of 219 · 165

Further left is better. Leaderboards measure different things: LMArena is people voting blind between two answers, the others are test scores. Click any row for the source.

## Best at · top three among current models

Replaced by DeepSeek-V4.1-Flash; it no longer leads any category among current models.

Where this model places in the top three of today's models, on category leaderboards and on our core tests.

## Facts and sources

Released

31 Jul 2026

Announcement

[api-docs.deepseek.com ↗](https://api-docs.deepseek.com/updates/)

Release entry

DeepSeek-V4-Flash (official release; 0731 weights)

Scored as

DeepSeek V4 Flash 0731

Artificial Analysis index

–

Output speed

–

Hugging Face

[deepseek-ai/DeepSeek-V4-Flash-0731 ↗](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731)

## DeepSeek releases · around this one

DeepSeek-V4-Flash (Apr 2026)

Open

24 APR 2026

index 145

DeepSeek-V4-Pro (Apr 2026)

Open

24 APR 2026

index 151

DeepSeek-V4-Flash

Open

31 JUL 2026

index 165

DeepSeek-V4-Pro

Open

13 AUG 2026

index 167

DeepSeek-V4-Flash-Vision-Exp

Open

21 AUG 2026

---
Source: https://themodelindex.org/models/deepseek-v4-flash-2026-07/ · The Model Index · data as of 2026-10-01