cyberneurova

A few things you should note as a Desktop User: all code and research chats are stored on your machine. This means we don't control your chats; when using the code section, all chats are stored on your local machine, and our agents are able to run tasks locally 👀. This is the best way for us to ensure user data and chats are secure. Your communication with the models is fully end-to-end and doesn't pass through any intermediary

Major bug fixes checkout the updates on the app

We are working to acquire a couple of Blackwell GPUs to run DeepSeek. Some models, like v8 versions, will be dropped and replaced with DeepSeek v4, which performs 120x better when writing complex code

V0.0.1 Beta

Desktop app linux and windows dropping today all in Beta

Our opensource abliterated deepseek-flash model just hit 77,000 downloads on huggingface showing the models quality

The web is shitty for complex tasks ⚠️ we are working on something 80x better. https://cyberneurova.ai

We will be deploying the DeepSeek v4 Pro Abliterated Version. This model will be accessible to paid users, with Pro users being prioritized. The model will not be deployed on the API yet, mainly due to compute power constraints and a high number of active subscribers. https://cyberneurova.ai

New Version just uploaded q2 and q8 Models will be added to https://cyberneurova.ai

Abliterated version of DeepSeek V4-Flash; this is the V2 version of what we released yesterday. Check out the benchmarks. The new version has advanced capabilities when it comes to writing or handling advanced tasks. It is one of the best tests so far, being able to write entire cyber-capable tools of 4K lines and evade Windows Antivirus.

This is the abliterated version of DeepSeek V4 Flash , created through ablation. We have open-sourced the V4-Flash version and quantized it from 160GB to 90GB. https://huggingface.co/cyberneurova/CyberNeurova-DeepSeek-V4-Flash-abliterated-GGUF

We are still 6500 GB VRAM short

We are buying 8k GB VRAM for 2weeks our budget is 6k$/ 1000 gb vram Mail us: admin@cyberneurova.ai

We have fixed the reasoning issue in the API; it should work well. The context for now is short, but we are expecting a 1 million context model to be ready by this week https://cyberneurova.ai