Sber presented Europe’s largest open-source project at AI Journey as it opened access to its flagship models — the GigaChat Ultra-Preview and Lightning, in addition to a new generation of the GigaAM-v3 open-source models for speech recognition and a full range of image and video generation models in the new Kandinsky 5.0 line, including the Video Pro, Video Lite and Image Lite.
The GigaChat Ultra-Preview, a new MoE model featuring 702 billion parameters, has been compiled specifically with the Russian language in mind and trained entirely from scratch. Read a detailed post from the team here.
For the first time in Russia, an MoE model of this scale has been trained entirely from scratch — without relying on any foreign weights. Training from scratch, and on such a scale to boot, is a challenge that few teams in the world have taken on.
Our flagship Kandinsky Video Pro model has caught up with Veo 3 in terms of visual quality and surpassed Wan 2.2-A14B. Read a detailed post from the team here.
The code and weights for all models are now available to all users under MIT license, including commercial use.
The GigaChat Ultra-Preview, a new MoE model featuring 702 billion parameters, has been compiled specifically with the Russian language in mind and trained entirely from scratch. Read a detailed post from the team here.
For the first time in Russia, an MoE model of this scale has been trained entirely from scratch — without relying on any foreign weights. Training from scratch, and on such a scale to boot, is a challenge that few teams in the world have taken on.
Our flagship Kandinsky Video Pro model has caught up with Veo 3 in terms of visual quality and surpassed Wan 2.2-A14B. Read a detailed post from the team here.
The code and weights for all models are now available to all users under MIT license, including commercial use.
AI Journey
AI Journey Conference on 19-21 November 2025. Key speakers in the area of artificial intelligence technology
AI Journey Conference on 19-21 November 2025. Key speakers in the area of artificial intelligence technology.
❤5
✅ Top 5 Mistakes to Avoid When Learning Java ❌☕
1️⃣ Skipping Core Concepts
Jumping into frameworks too early is a trap. Master variables, data types, loops, conditionals, and OOP (Classes, Objects, Inheritance, Polymorphism) first.
2️⃣ Ignoring the Java Ecosystem
Java isn’t just the language. Understand the JDK, JVM, and JRE. Know how Java compiles and runs code.
3️⃣ Not Using an IDE Properly
Using basic editors slows you down. Learn IntelliJ IDEA or Eclipse. Explore features like debugging, refactoring, and code suggestions.
4️⃣ Avoiding Error Handling
Try-catch blocks and exception handling are core in Java. Skipping them leads to messy and unstable code.
5️⃣ Neglecting Practical Projects
Only reading theory won’t help. Build Java apps like a to-do list, calculator, or REST API using Spring Boot to apply what you’ve learned.
Additional Common Mistakes Found in 2025 Experts' Advice:
⦁ Comparing strings with
⦁ Not checking for
⦁ Writing all code inside the
⦁ Overcomplicating simple problems with unnecessary patterns
⦁ Not following proper naming conventions harms code clarity
💬 Tap ❤️ for more!
1️⃣ Skipping Core Concepts
Jumping into frameworks too early is a trap. Master variables, data types, loops, conditionals, and OOP (Classes, Objects, Inheritance, Polymorphism) first.
2️⃣ Ignoring the Java Ecosystem
Java isn’t just the language. Understand the JDK, JVM, and JRE. Know how Java compiles and runs code.
3️⃣ Not Using an IDE Properly
Using basic editors slows you down. Learn IntelliJ IDEA or Eclipse. Explore features like debugging, refactoring, and code suggestions.
4️⃣ Avoiding Error Handling
Try-catch blocks and exception handling are core in Java. Skipping them leads to messy and unstable code.
5️⃣ Neglecting Practical Projects
Only reading theory won’t help. Build Java apps like a to-do list, calculator, or REST API using Spring Boot to apply what you’ve learned.
Additional Common Mistakes Found in 2025 Experts' Advice:
⦁ Comparing strings with
== instead of .equals() causes bugs⦁ Not checking for
null leads to NullPointerException⦁ Writing all code inside the
main method becomes messy⦁ Overcomplicating simple problems with unnecessary patterns
⦁ Not following proper naming conventions harms code clarity
💬 Tap ❤️ for more!
❤10😁1
Tired of AI that refuses to help?
@UnboundGPT_bot doesn't lecture. It just works.
✓ Multiple models (GPT-4o, Gemini, DeepSeek)
✓ Image generation & editing
✓ Video creation
✓ Persistent memory
✓ Actually uncensored
Free to try → @UnboundGPT_bot or https://ko2bot.com
@UnboundGPT_bot doesn't lecture. It just works.
✓ Multiple models (GPT-4o, Gemini, DeepSeek)
✓ Image generation & editing
✓ Video creation
✓ Persistent memory
✓ Actually uncensored
Free to try → @UnboundGPT_bot or https://ko2bot.com
Ko2Bot
Ko2 - Advanced AI Platform
Ko2 - Multi-model AI platform with GPT-4o, Claude, Gemini, DeepSeek for text and FLUX, Grok, Qwen for image generation.
❤2