Quick Answer: DeepSeek V4 Flash is an efficiency-focused AI model from Chinese AI startup DeepSeek, first released as a preview on April 23-24, 2026, and formally released as DeepSeek-V4-Flash-0731 on July 31, 2026. It is a Mixture-of-Experts model with 284 billion total parameters, 13 billion of which activate per token, a 1-million-token context window, and MIT-licensed open weights available for commercial use.

What Happened?

DeepSeek released the official version of its V4-Flash model on July 31, 2026, retraining the April preview version with a new post-training pipeline focused on coding, autonomous agent tasks, reasoning and tool use, while keeping the same underlying architecture and parameter count.

Why Is This Trending?

The model is drawing search interest for combining strong benchmark performance with unusually low API pricing, positioning it as a low-cost alternative to larger proprietary models from U.S. AI labs, intensifying competition in the global AI market.

Key Details

• Architecture: Mixture-of-Experts (MoE).
• Total parameters: 284 billion; activated parameters per token: 13 billion.
• Context window: 1 million tokens.
• License: MIT, allowing commercial and on-premise deployment.
• API pricing: approximately $0.14 per million input tokens and $0.28 per million output tokens.
• DeepSeek reports the model outperforms its own V4-Pro preview and GLM-5.2 on published agent benchmarks.

What This Means for Developers

The combination of open MIT-licensed weights and low API costs makes V4 Flash appealing for developers building high-throughput applications who want strong reasoning and coding performance without the cost of larger proprietary models.

What We Know So Far

Confirmed: the model’s specifications, license, release timeline and DeepSeek’s own published benchmark comparisons. Not independently verified here: third-party, apples-to-apples benchmark comparisons against every competing model, since DeepSeek’s benchmark figures come from the company’s own reporting.

Frequently Asked Questions

Is DeepSeek V4 Flash free to use?

The model weights are open-sourced under the MIT license, meaning developers can self-host it, while DeepSeek’s hosted API charges per token.

When was DeepSeek V4 Flash released?

A preview version launched April 23-24, 2026, with the official DeepSeek-V4-Flash-0731 release on July 31, 2026.

How big is the model?

It has 284 billion total parameters, with 13 billion activated per token, and supports a 1-million-token context window.

Can businesses use it commercially?

Yes, the MIT license permits commercial and on-premise deployment without access restrictions.

Key Takeaways

• DeepSeek V4 Flash is an open-source, MIT-licensed AI model.
• It was officially released on July 31, 2026, as DeepSeek-V4-Flash-0731.
• It has 284 billion total parameters with 13 billion active per token.
• It supports a 1-million-token context window at low API cost.