The Silent Revolution: DeepSeek V3.1 Is Here (and Free!) 🤯

Breaking news from the AI world! DeepSeek, the Chinese AI that garnered significant attention this year, has just released a new version: DeepSeek V3.1. The most remarkable aspect? This powerful update is completely free to use directly within the chat interface. What makes it even more intriguing is that the release was quietly rolled out, without an official announcement on Twitter, but through Hugging Face and DeepSeek's official WeChat group.

What Makes DeepSeek V3.1 So Special?

DeepSeek V3.1 promises significant improvements over its predecessors and competing models:

🚀 Impressive Performance

On SVG Bench, a leading benchmark platform, DeepSeek V3.1 scores 53.1%, which is higher than many top models such as Gemini 2.5 Flash, GP5 Chat, A04 Mini Flash Light, and Kim K2. It also scores 71.6% on ADA.

📈 Doubled Context Window

The context window has been doubled from 64K to 128K tokens (compared to DeepSeek V3). While still behind some models with 1 million tokens, this is a substantial advancement that allows for processing much larger documents and maintaining context over longer conversations.

⚡ Blazingly Fast and Efficient

The new version is significantly faster, up to 2.5 times faster in thinking time than its predecessor. Furthermore, DeepSeek V3.1 is 68 times cheaper and 1% more efficient than Opus. The usage cost is described as "super cheap," and there appears to be no limit to its free use in the chat.

🔄 Hybrid Architecture and "Auto-Mode"

DeepSeek V3.1 introduces a hybrid architecture that supports "thinking and search". Previously, users had to manually switch between "deep think" and "non-deep think" modes. Now, the model seems to switch automatically, similar to the approach seen in models like GPT-5.

🔓 Open-Source Availability

DeepSeek V3.1 is downloadable on Hugging Face under an open-source license. This aligns with the growing trend of open-source AI models, making advanced AI technology more accessible to researchers and developers worldwide.

📚 Up-to-Date Knowledge

The AI itself indicates that its knowledge is up-to-date as of July 2024, ensuring users have access to relatively recent information.

🎯 Model Size

The model has 685 billion parameters, positioning it among the largest and most capable language models available today.

How to Access DeepSeek V3.1?

To use DeepSeek V3.1, you simply need to go to chat.deepseek.com and sign up. You can start chatting directly and benefit from the new features. While there are no official API updates yet on platforms like OpenRouter for V3.1, and no local version of V3.1 has been released, you can fully utilize the free chat environment.

For advanced users who wish to run DeepSeek locally, it is possible to do so with Lama and Browser Use Web UI. This allows you to have DeepSeek control your laptop for fully autonomous computing tasks. Within the chat environment, you can preview HTML outputs directly or download and host them on Netlify.

Is DeepSeek V3.1 Also Good for Coding?

The sources provide mixed signals regarding DeepSeek V3.1's coding capabilities. While it is capable of generating code (such as a Pong game in HTML or Python), it failed in a direct comparison with Claude Opus 4.8 and Gemini when attempting to create a playable game. Gemini 2.5 Pro is still considered the best for coding tasks and performed better than DeepSeek V3.1 in tests.

DeepSeek V3.1 is excellent for general chat purposes, but for more complex coding tasks, it still seems to have some catching up to do.

The Future of Open-Source AI

DeepSeek V3.1 contributes to the rapidly growing open-source AI movement, alongside models like Google Gemma and Mistral. The trend is clear: open-source AI is at its peak. The ability to run powerful models locally on devices such as phones is seen as the future of AI, an area where Google currently appears to be leading.

Where to Find More Information?

This is just the beginning for DeepSeek V3.1, and its potential is immense. For those who wish to delve deeper and learn all the ins and outs, there is an extensive course and detailed SOPs (Standard Operating Procedures) available in the AI Profit Boardroom. This community offers daily updates, coaching calls, and an active environment for anyone looking to stay current with the latest AI developments.

Stay tuned, as the world of AI is evolving faster than ever!