What's new

Our chat now features GPT-5 - OpenAI's next-generation flagship artificial intelligence model! Released on August 7, 2025, this model represents a quantum leap in AI technology development 🎉

Key Features 🎯

GPT-5 impresses with its technical specifications:

• Unified Architecture 🏗️ - combines advanced reasoning, multimodal input, and task execution into a single system
• Enhanced Multimodality 🖼️ - supports text, images, audio, and potentially video with an input limit of 272,000 tokens
• Improved Memory 🧠 - effectively retains information across sessions for continuous project work
• MoE Architecture ⚡ - dynamic allocation of computational resources for specialized tasks
• Autonomous Tools 🤖 - performs actions without constant user input

Performance and Achievements 📊

The model demonstrates outstanding results:

• 94.6% on AIME 2025 🧮 - new record in mathematical problems
• 74.9% on SWE-bench Verified 💻 - excellence in real-world coding
• 88% on Aider Polyglot 🌐 - multilingual programming
• 80% reduction in hallucinations ✅ - compared to previous models
• 45% fewer factual errors 📚 - than GPT-4o when working with web search

Available Models 🎭

GPT-5 is available in two variants:

• Regular - full-featured version
• Mini - optimized for speed

Each model can operate at four reasoning levels: minimal, low, medium, and high.

Availability 🌟

GPT-5 is already integrated into our chat and ready to help you with:

• Complex multi-step workflows 🔄
• Writing and editing texts 📝
• Programming and code debugging 🛠️
• Visual analysis and understanding 👁️
• Scientific and academic research 🔬
• Creative projects and idea generation 💡

Try the power of GPT-5 right now in our chat! 🚀

We are excited to announce the addition of Grok-4 - the latest flagship AI model from xAI to our chat. This revolutionary model, released in July 2025, represents a significant breakthrough in the field of artificial intelligence.

Key Features

Grok-4 boasts impressive technical specifications:

• Extended context window - support for over 130,000 tokens for working with large documents and complex tasks
• Multimodal capabilities - analysis of text, images, and real-time visual content understanding
• Built-in programming support - specialized capabilities for development and code debugging
• Real-time search integration - access to up-to-date information from the internet
• Advanced tool usage - trained to independently select and apply various tools for problem-solving

Performance and Achievements

The model demonstrates outstanding results:

• Superiority over PhD level in all academic disciplines without exception
• 44.4% on Humanity's Last Exam - significant superiority over competitors
• 15.9% on ARC-AGI V2 - nearly double improvement compared to previous models

Availability

Grok-4 is already integrated into our chat and ready to help you with:

• Complex analysis and research
• Programming and development
• Creative tasks
• Working with images and multimedia
• Solving academic and professional questions

Key Features

ImageGPT is a powerful model for creating images based on text descriptions with the following advantages:

• High-precision prompt following - the model accurately reproduces even complex descriptions with multiple elements
• Enhanced transparent background handling - perfect for creating logos, stickers, and interface elements
• Complex scene processing - capable of correctly generating images with dozens of various objects and colors
• Artistic expressiveness - combines technical precision with a creative approach

Applications

The model is especially effective for:

• Creating logos and branding materials
• Generating stickers and graphic elements
• Developing user interface elements
• Creating illustrations for content

ImageGPT is already available for free use and ready to help implement your creative ideas with unprecedented accuracy and quality.

Mistral AI has released a new language model, Mistral Medium 3. It provides high performance comparable to Claude Sonnet 3.7 but is 8 times cheaper.
The model outperforms competitors in coding and STEM tasks, supports multimodality, and has a context window of 128,000 tokens.

Google has released an updated version of its flagship model, Gemini 2.5 Pro Preview (05-06). Key improvements include program code generation,
web application creation, and video analysis. The model topped the WebDev Arena Leaderboard and scored 84.8% on the VideoMME test.

🚀 Awesome news from OpenAI: The GPT-4.1 model family is out! 🚀

OpenAI has announced a new generation of AI: GPT-4.1, GPT-4.1 mini

What has changed compared to GPT-4o?

• ✅ Programming on a new level: GPT-4.1 significantly outperforms previous versions in coding tasks.
• ✅ Better understanding: The models follow complex instructions more accurately and maintain context in long dialogues.
• ✅ Massive context: models support up to 1 MILLION tokens! This allows analyzing huge volumes of code or documents in full.
• ✅ Up-to-date knowledge: The knowledge base is updated through June 2024.
• ✅ Efficiency and cost: The new models run faster and cost less! For example, GPT-4.1 mini matches GPT-4o’s intelligence level but is almost twice as fast and 83% (!) cheaper.

What is the difference between the new models?

• GPT-4.1: The flagship for the most complex tasks (coding, research).
• GPT-4.1 mini: The optimal balance of power, speed, and cost. Supports text and images.

Now the result size can be up to 4096*4096 pixels, provided that the mask size does not exceed 30% of the image size.

Now, in the task history, you can see the current progress of processing long video files.

This error occurred during batch processing of images (more than 8 images).

New interface:

• You can change a face in any photo.
• Face selection feature (serial number).
• Improved face quality and increased output resolution.

Google has released Gemini 2.5 Pro — a powerful AI that handles tasks related to logic, mathematics, and coding. Its capabilities were assessed using challenging tests, such as GPQA — a set of graduate-level questions in physics, chemistry, and biology, and AIME — a mathematics competition for high school students. These tests demonstrate how well the AI solves problems that are difficult even for experts.