Model intelligence / Live profile

DeepSeek V4 Flash

Track DeepSeek V4 Flash from OpenAI Group PBC across launches, availability, pricing, benchmarks, capabilities, and consequential updates. This page updates automatically as substantive Superpower Daily coverage is published.

Official page

Coverage telemetryActive

Developer / company

OpenAI Group PBC

Published3

Launches1

Pricing0

Benchmarks1

Latest coverageDeepSeek Gives V4-Flash a Separate Vision API With a 384-Token Image Cap

Published coverage3 stories

Last 90 days3 updates

Coverage dates3 days

Profile statusIndexed

Releasedeepseek-v4-flash-vision-exp

AvailabilityAvailable through Chat Completions, Messages, and Responses requests

ContextNot disclosed

Input / 1M$0.22

Output / 1M$0.66

Latest coverage

What changed around DeepSeek V4 Flash.

Browse all models

1 hour agoModels

BenchmarkModels

LLMScholarBench Tests 22 Models and Finds an Accuracy-Representation Trade-Off

The benchmark treats AI-generated expert lists as recommender systems to audit, showing that factual correctness and balanced representation cannot yet be improved together reliably.

3 min

Aug 21, 2026Business

Platform shiftBusiness

China’s New AI Hub Has 12.5 Gigawatts Planned—and a Water Problem

Cold weather, cheap electricity and faster fiber have made Ulanqab viable for both AI training and user-facing inference. Its growth now depends on whether the region can manage water pressure and reduce its reliance on coal.

3 min

Durable signals

Launches, pricing, and proof.

011 records

Launches and releases

DeepSeek Gives V4-Flash a Separate Vision API With a 384-Token Image CapAug 22, 2026

020 records

Pricing and access

No material pricing update yet.

031 records

Benchmarks and evaluations

LLMScholarBench Tests 22 Models and Finds an Accuracy-Representation Trade-Off1 hour ago

Update timeline

The maintained record.

LaunchAug 22, 2026DeepSeek Gives V4-Flash a Separate Vision API With a 384-Token Image CapThe experimental model gives developers several familiar ways to submit images and a fixed upper bound on image-token use. It leaves the harder questions—visual accuracy, latency and reliability—for users to test.Platform shiftAug 21, 2026China’s New AI Hub Has 12.5 Gigawatts Planned—and a Water ProblemCold weather, cheap electricity and faster fiber have made Ulanqab viable for both AI training and user-facing inference. Its growth now depends on whether the region can manage water pressure and reduce its reliance on coal.Model launchesAug 20, 2026DeepSeek launches experimental deepseek-v4-flash-vision-exp vision modelDeepSeek announced an experimental vision model supporting Chat Completions, Messages, and Responses requests, with a 384-token billing ceiling per image.Model launchesAug 20, 2026DeepSeek releases V4-Flash-Vision-Exp experimental vision modelDeepSeek released V4-Flash-Vision-Exp, an experimental multimodal model available through its API and compatible with OpenAI and Anthropic API formats.

Your watchlist