Agent upgrade, build 0731

DeepSeek V4 Flash online, pay per use

DeepSeek V4 Flash is the cheap half of the V4 family: about 284B total parameters with only 13B active per token, a 1 million token context window and hybrid attention built for long agent runs. The 0731 build, released on July 31, 2026, kept the same body and redid the post-training, and DeepSeek reports it now beats its own V4-Pro preview across nine agent benchmarks. On avots.ai you run it pay per use, with no DeepSeek account and no subscription, and the same balance also covers Claude Opus 5, GPT-5.6, Gemini and 40+ image, video and music models.

Use DeepSeek V4 Flash now ✦ See pricing
Specs

DeepSeek V4 Flash at a glance

284B total, 13B active

A sparse mixture of experts: only about 13B parameters fire per token, which is why it answers fast and costs so little.

1M token context

A full million tokens in one conversation, with very large outputs, so whole repositories and document sets fit in a single pass.

0.14 USD in, 0.28 USD out

Per million tokens on DeepSeek's own list, with cached input at 0.0028 USD, so repeated context in agent loops is almost free.

82.7 Terminal-Bench 2.1

DeepSeek reported for the 0731 build, alongside 70.3 on Toolathlon verified and 76.7 on Cybergym.

Index score 50

On the independent Artificial Analysis Intelligence Index, against a median of 25 for comparable models.

Text in, text out

No image input on this one. For screenshots and scans switch to Qwen3.7 or Gemini on the same balance.

Real numbers

V4 Flash next to V4-Pro and Qwen3.7 Flash

 DeepSeek V4 FlashDeepSeek V4-ProQwen3.7 Flash
Price per 1M in / out$0.14 / $0.28$0.435 / $0.87$0.03 / $0.13
Parameters284B total, 13B activeabout 1.6T total, 49B activenot published
Context window1M tokens1M tokens1M tokens
Terminal-Bench 2.182.7not publishednot published
Toolathlon verified70.3not publishednot published
Image inputnonoyes
On avots.aipay per usepay per usepay per use

Agent scores are vendor reported and very sensitive to the harness, so treat them as a direction, not a verdict. What is not in doubt is the price gap: Flash runs at roughly a third of V4-Pro per token, and Qwen3.7 Flash undercuts both while adding image input. All three sit in the same picker.

What is new

What the 0731 build changed

Same body, new training

Identical architecture and size to the preview build. Everything gained came from redone post-training, not a new base model.

Agent behaviour first

DeepSeek reports 54.4 on DeepSWE, 54.2 on NL2Repo and 25.2 on Agents' Last Exam, its strongest agent set yet.

Cheap repeated context

Cache hit input at 0.0028 USD per million tokens makes tool loops that resend the same prompt genuinely cheap.

Effort levels

High and maximum reasoning effort are supported, so you spend more thinking only on the calls that need it.

Try V4 Flash on avots
Three steps

How to use DeepSeek V4 Flash

Create an account

Sign up on app.avots.ai with email, Google or Telegram. Takes a minute.

Top up from 0.99 EUR

1,000 tokens cost 0.99 EUR. No plan to choose, no renewal, the balance sits there until you use it.

Pick V4 Flash

Choose DeepSeek V4 Flash in the model picker, or let Agent Avots route for you. Every reply shows its token cost.

Sign up and start
One balance

Where you can use DeepSeek V4 Flash

Web app

app.avots.ai on desktop and mobile, no install needed.

Telegram

V4 Flash in @AvotsAIbot, same balance, from any phone.

MCP

Use it inside Claude Desktop, Cursor or Cline via the MCP server.

Try DeepSeek V4 Flash for under a euro

1,000 tokens cost 0.99 EUR and a Flash answer costs a handful of them. The same balance unlocks Claude Opus 5, GPT-5.6, Gemini, Veo video and Nano Banana images, with no subscription anywhere.

Sign up and chat ✦
Guides

Related pages

DeepSeek V4

The full family, V4-Pro and V4-Flash, open weights and 1M context.

Qwen3.7

Alibaba's Max, Plus and Flash tiers, 1M context, vision on the small ones.

Kimi K3

Moonshot's frontier open model, number one at coding, 1M context.

Claude Opus 5

Anthropic's newest flagship, 1M context, 96% SWE-bench Verified.

avots.ai runs 40+ AI models for chat, image, video and audio on one balance. See the full list on the Tools page or open the app and start creating.

Browse all model pages Open the app
FAQ

DeepSeek V4 Flash questions

What is DeepSeek V4 Flash?
DeepSeek V4 Flash is the small, fast member of the DeepSeek V4 family: a mixture of experts model with about 284B total parameters and only 13B active per token, a 1 million token context window and text input and output. It is built for coding assistants, tool calling and long agent loops where speed and price matter more than absolute peak reasoning.
What changed in the 0731 build?
DeepSeek released the V4-Flash-0731 build on July 31, 2026. The architecture and size did not change, only the post-training was redone, and the gains are in agent behaviour and tool calling. DeepSeek reports 82.7 on Terminal-Bench 2.1, 70.3 on Toolathlon verified and 76.7 on Cybergym for that build.
How much does DeepSeek V4 Flash cost?
DeepSeek lists 0.14 USD per million input tokens and 0.28 USD per million output tokens, with cached input at 0.0028 USD. The independent Artificial Analysis index puts its blended cost near 0.06 USD per million tokens. On avots.ai you do not pay per million: you top up from 0.99 EUR for 1,000 avots tokens and every reply shows its cost before it runs.
Flash or V4-Pro, which should I pick?
Pick Flash for volume: agent loops, refactors across many files, chat that has to feel instant, anything where you send the same long context again and again. Pick V4-Pro for the hardest single answers, where its 1.6T parameter body and 93.5 LiveCodeBench score earn the higher price. Both sit in the same picker on one balance, so you can switch mid task.
Does DeepSeek V4 Flash read images?
No. V4 Flash takes text in and returns text out. If you need to send screenshots or scans, switch to Qwen3.7, Gemini or GPT-5.6 in the same picker. One avots.ai balance covers all of them plus image, video and music models.
Try it on avots ✦
Start creating