GigaChat 3.5 Ultra Publicly Released — The New Generation of the Flagship Model
What’s inside:
🔘 A proprietary hybrid MLA + Gated DeltaNet architecture with a dedicated stabilization framework, without which this hybrid setup would not train reliably at this scale;
🔘 Gated Attention: the model can locally down-weight overly strong signals from the attention layer;
🔘 GatedNorm: normalization with an explicit gate that controls signal magnitude across features;
🔘 Approximately 4x lower KV cache per token: with the same memory budget, the model can support 2.14x longer context and deliver a 20% throughput increase under load;
🔘 Two MTP heads, enabling up to 2.2x faster generation;
🔘 FP8 across all training stages with no quality degradation compared with bf16, enabled by custom Triton and CUDA kernels;
🔘 A new online RL stage after SFT and DPO.
Results:
🔘 GigaChat-3.5-Ultra-Base outperforms DeepSeek V3.2 Exp Base and DeepSeek V4 Flash Base on average across a set of general, math, and code benchmarks:
🔘 GigaChat-3.5-Ultra-Instruct is comparable to DeepSeek V3.2 in terms of average score, despite having half the size;
🔘 According to the MiniMax-M2.7 LLM judge, the average win rate against GigaChat 3.1 Ultra is 75.9%, and against GPT-5 is 68.7%.
➡️ HuggingFace
The GigaChat team has released GigaChat 3.5 Ultra as open source—a new 432B model under the MIT license. This is the first open-source hybrid of GatedDeltaNet and MLA scaled to hundreds of billions of parameters, featuring a proprietary training recipe we refined through more than 1,500 experiments. The model has grown in terms of code, mathematics, agent scenarios, and application domains—yet it’s 40% smaller than GigaChat 3.1 Ultra.
What’s inside:
Results:
The entire stack — data (our own LLM-filtered Common Crawl, 600+ programming languages in the code), architecture, training methodology, and infrastructure — was built end-to-end by GigaChat team.
Please open Telegram to view this post
VIEW IN TELEGRAM
❤3
𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 𝗙𝗥𝗘𝗘 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗥𝗲𝘀𝗼𝘂𝗿𝗰𝗲𝘀🎓
Offers a wide range of free learning resources through Microsoft Learn, helping students, freshers, and professionals build job-ready skills at their own pace.
✅ 100% FREE self-paced learning modules
✅ Official learning platform from Microsoft
🔗 𝗘𝗻𝗿𝗼𝗹𝗹 𝗙𝗼𝗿 𝗙𝗥𝗘𝗘👇:
https://pdlink.in/4paqRJS
Explore Microsoft’s free resources. Build in-demand skills and make your profile stronger.
Offers a wide range of free learning resources through Microsoft Learn, helping students, freshers, and professionals build job-ready skills at their own pace.
✅ 100% FREE self-paced learning modules
✅ Official learning platform from Microsoft
🔗 𝗘𝗻𝗿𝗼𝗹𝗹 𝗙𝗼𝗿 𝗙𝗥𝗘𝗘👇:
https://pdlink.in/4paqRJS
Explore Microsoft’s free resources. Build in-demand skills and make your profile stronger.
❤2
✅ If you're serious about learning Python for data science, automation, or interviews — just follow this roadmap 🐍💻
1. Install Python Jupyter Notebook (via Anaconda or VS Code)
2. Learn print(), variables, and data types 📦
3. Understand lists, tuples, sets, and dictionaries 🔁
4. Master conditional statements (if, elif, else) ✅❌
5. Learn loops (for, while) 🔄
6. Functions – defining and calling functions 🔧
7. Exception handling – try, except, finally ⚠️
8. String manipulations formatting ✂️
9. List dictionary comprehensions ⚡
10. File handling (read, write, append) 📁
11. Python modules packages 📦
12. OOP (Classes, Objects, Inheritance, Polymorphism) 🧱
13. Lambda, map, filter, reduce 🔍
14. Decorators Generators ⚙️
15. Virtual environments pip installs 🌐
16. Automate small tasks using Python (emails, renaming, scraping) 🤖
17. Basic data analysis using Pandas NumPy 📊
18. Explore Matplotlib Seaborn for visualization 📈
19. Solve Python coding problems on LeetCode/HackerRank 🧠
20. Watch a mini Python project (YouTube) and build it step by step 🧰
21. Pick a domain (web dev, data science, automation) and go deep 🔍
22. Document everything on GitHub 📁
23. Add 1–2 real projects to your resume 💼
Trick: Copy each topic above, search it on YouTube, watch a 10-15 min video, then code along.
🎯 This method builds actual understanding + project experience for interviews!
💬 Tap ❤️ for more!
1. Install Python Jupyter Notebook (via Anaconda or VS Code)
2. Learn print(), variables, and data types 📦
3. Understand lists, tuples, sets, and dictionaries 🔁
4. Master conditional statements (if, elif, else) ✅❌
5. Learn loops (for, while) 🔄
6. Functions – defining and calling functions 🔧
7. Exception handling – try, except, finally ⚠️
8. String manipulations formatting ✂️
9. List dictionary comprehensions ⚡
10. File handling (read, write, append) 📁
11. Python modules packages 📦
12. OOP (Classes, Objects, Inheritance, Polymorphism) 🧱
13. Lambda, map, filter, reduce 🔍
14. Decorators Generators ⚙️
15. Virtual environments pip installs 🌐
16. Automate small tasks using Python (emails, renaming, scraping) 🤖
17. Basic data analysis using Pandas NumPy 📊
18. Explore Matplotlib Seaborn for visualization 📈
19. Solve Python coding problems on LeetCode/HackerRank 🧠
20. Watch a mini Python project (YouTube) and build it step by step 🧰
21. Pick a domain (web dev, data science, automation) and go deep 🔍
22. Document everything on GitHub 📁
23. Add 1–2 real projects to your resume 💼
Trick: Copy each topic above, search it on YouTube, watch a 10-15 min video, then code along.
🎯 This method builds actual understanding + project experience for interviews!
💬 Tap ❤️ for more!
❤16
✅ Statistics & Probability Cheatsheet 📚🧠
📌 Descriptive Statistics:
⦁ Mean = (Σx) / n
⦁ Median = Middle value
⦁ Mode = Most frequent value
⦁ Variance (σ²) = Σ(x - μ)² / n
⦁ Std Dev (σ) = √Variance
⦁ Range = Max - Min
⦁ IQR = Q3 - Q1
📌 Probability Basics:
⦁ P(A) = Outcomes A / Total Outcomes
⦁ P(A ∩ B) = P(A) × P(B) (if independent)
⦁ P(A ∪ B) = P(A) + P(B) - P(A ∩ B)
⦁ Conditional: P(A|B) = P(A ∩ B) / P(B)
⦁ Bayes’ Theorem: P(A|B) = [P(B|A) × P(A)] / P(B)
📌 Common Distributions:
⦁ Binomial (fixed trials)
⦁ Normal (bell curve)
⦁ Poisson (rare events over time)
⦁ Uniform (equal probability)
📌 Inferential Stats:
⦁ Z-score = (x - μ) / σ
⦁ Central Limit Theorem: sampling dist ≈ Normal
⦁ Confidence Interval: CI = x ± z*(σ/√n)
📌 Hypothesis Testing:
⦁ H₀ = No effect; H₁ = Effect present
⦁ p-value < α → Reject H₀
⦁ Tests: t-test (small samples), z-test (known σ), chi-square (categorical data)
📌 Correlation:
⦁ Pearson: linear relation (–1 to 1)
⦁ Spearman: rank-based correlation
🧪 Tools to Practice:
Python packages:
Visualization:
💡 Quick tip: Use these formulas to crush interviews and build solid ML foundations!
💬 Tap ❤️ for more
📌 Descriptive Statistics:
⦁ Mean = (Σx) / n
⦁ Median = Middle value
⦁ Mode = Most frequent value
⦁ Variance (σ²) = Σ(x - μ)² / n
⦁ Std Dev (σ) = √Variance
⦁ Range = Max - Min
⦁ IQR = Q3 - Q1
📌 Probability Basics:
⦁ P(A) = Outcomes A / Total Outcomes
⦁ P(A ∩ B) = P(A) × P(B) (if independent)
⦁ P(A ∪ B) = P(A) + P(B) - P(A ∩ B)
⦁ Conditional: P(A|B) = P(A ∩ B) / P(B)
⦁ Bayes’ Theorem: P(A|B) = [P(B|A) × P(A)] / P(B)
📌 Common Distributions:
⦁ Binomial (fixed trials)
⦁ Normal (bell curve)
⦁ Poisson (rare events over time)
⦁ Uniform (equal probability)
📌 Inferential Stats:
⦁ Z-score = (x - μ) / σ
⦁ Central Limit Theorem: sampling dist ≈ Normal
⦁ Confidence Interval: CI = x ± z*(σ/√n)
📌 Hypothesis Testing:
⦁ H₀ = No effect; H₁ = Effect present
⦁ p-value < α → Reject H₀
⦁ Tests: t-test (small samples), z-test (known σ), chi-square (categorical data)
📌 Correlation:
⦁ Pearson: linear relation (–1 to 1)
⦁ Spearman: rank-based correlation
🧪 Tools to Practice:
Python packages:
scipy.stats, statsmodels, pandas Visualization:
seaborn, matplotlib💡 Quick tip: Use these formulas to crush interviews and build solid ML foundations!
💬 Tap ❤️ for more
❤23👌1
Data is the fuel but AI is the Machinery.
The people who know how to use both will lead the future.
Become one with TiHAN IIT Hyderabad's AI & ML Program.
✅ Learn live from TiHAN scientists, IIT professors & industry experts
✅ Build hands-on projects with Flipkart & Mamaearth
✅ Assured interview at TiHAN IIT Hyderabad with 9+ CGPA
✅ Placement support across 5000+ companies through Masai
Online Entrance Exam: 19th July
🔗 Register: https://tinyurl.com/datasimplifier-17jul-tihan-006
The people who know how to use both will lead the future.
Become one with TiHAN IIT Hyderabad's AI & ML Program.
✅ Learn live from TiHAN scientists, IIT professors & industry experts
✅ Build hands-on projects with Flipkart & Mamaearth
✅ Assured interview at TiHAN IIT Hyderabad with 9+ CGPA
✅ Placement support across 5000+ companies through Masai
Online Entrance Exam: 19th July
🔗 Register: https://tinyurl.com/datasimplifier-17jul-tihan-006
❤3
Machine Learning & Artificial Intelligence | Data Science Free Courses
Data is the fuel but AI is the Machinery. The people who know how to use both will lead the future. Become one with TiHAN IIT Hyderabad's AI & ML Program. ✅ Learn live from TiHAN scientists, IIT professors & industry experts ✅ Build hands-on projects with…
Final 6 Hours Left!
To register for TiHAN IIT Hyderabad's AI & ML Program.
Don't miss your chance to:
• Learn from India's best scientists at TiHAN, IIT Professors and industry experts
• Direct Interview at TiHAN IIT Hyderabad with 9+ CGPA
Register before the Admission Closes!
To register for TiHAN IIT Hyderabad's AI & ML Program.
Don't miss your chance to:
• Learn from India's best scientists at TiHAN, IIT Professors and industry experts
• Direct Interview at TiHAN IIT Hyderabad with 9+ CGPA
Register before the Admission Closes!
❤2
Preparing for an SQL Interview? Here’s What You Need to Know!
If you’re aiming for a data-related role, strong SQL skills are a must.
Basics:
→ Learn about the difference between SQL and MySQL, primary keys, foreign keys, and how to use JOINs.
Intermediate:
→ Get into more detailed topics like subqueries, views, and how to use aggregate functions like COUNT and SUM.
Advanced:
→ Explore more complex ideas like window functions, transactions, and optimizing SQL queries for better performance.
🡲 Quick Tip: Practice writing these queries and explaining your thought process.
If you’re aiming for a data-related role, strong SQL skills are a must.
Basics:
→ Learn about the difference between SQL and MySQL, primary keys, foreign keys, and how to use JOINs.
Intermediate:
→ Get into more detailed topics like subqueries, views, and how to use aggregate functions like COUNT and SUM.
Advanced:
→ Explore more complex ideas like window functions, transactions, and optimizing SQL queries for better performance.
🡲 Quick Tip: Practice writing these queries and explaining your thought process.
👍1
Machine Learning – Essential Concepts 🚀
1️⃣ Types of Machine Learning
Supervised Learning – Uses labeled data to train models.
Examples: Linear Regression, Decision Trees, Random Forest, SVM
Unsupervised Learning – Identifies patterns in unlabeled data.
Examples: Clustering (K-Means, DBSCAN), PCA
Reinforcement Learning – Models learn through rewards and penalties.
Examples: Q-Learning, Deep Q Networks
2️⃣ Key Algorithms
Regression – Predicts continuous values (Linear Regression, Ridge, Lasso).
Classification – Categorizes data into classes (Logistic Regression, Decision Tree, SVM, Naïve Bayes).
Clustering – Groups similar data points (K-Means, Hierarchical Clustering, DBSCAN).
Dimensionality Reduction – Reduces the number of features (PCA, t-SNE, LDA).
3️⃣ Model Training & Evaluation
Train-Test Split – Dividing data into training and testing sets.
Cross-Validation – Splitting data multiple times for better accuracy.
Metrics – Evaluating models with RMSE, Accuracy, Precision, Recall, F1-Score, ROC-AUC.
4️⃣ Feature Engineering
Handling missing data (mean imputation, dropna()).
Encoding categorical variables (One-Hot Encoding, Label Encoding).
Feature Scaling (Normalization, Standardization).
5️⃣ Overfitting & Underfitting
Overfitting – Model learns noise, performs well on training but poorly on test data.
Underfitting – Model is too simple and fails to capture patterns.
Solution: Regularization (L1, L2), Hyperparameter Tuning.
6️⃣ Ensemble Learning
Combining multiple models to improve performance.
Bagging (Random Forest)
Boosting (XGBoost, Gradient Boosting, AdaBoost)
7️⃣ Deep Learning Basics
Neural Networks (ANN, CNN, RNN).
Activation Functions (ReLU, Sigmoid, Tanh).
Backpropagation & Gradient Descent.
8️⃣ Model Deployment
Deploy models using Flask, FastAPI, or Streamlit.
Model versioning with MLflow.
Cloud deployment (AWS SageMaker, Google Vertex AI).
Data Science Resources
👇👇
https://whatsapp.com/channel/0029Va4QUHa6rsQjhITHK82y
Like for more 😄
1️⃣ Types of Machine Learning
Supervised Learning – Uses labeled data to train models.
Examples: Linear Regression, Decision Trees, Random Forest, SVM
Unsupervised Learning – Identifies patterns in unlabeled data.
Examples: Clustering (K-Means, DBSCAN), PCA
Reinforcement Learning – Models learn through rewards and penalties.
Examples: Q-Learning, Deep Q Networks
2️⃣ Key Algorithms
Regression – Predicts continuous values (Linear Regression, Ridge, Lasso).
Classification – Categorizes data into classes (Logistic Regression, Decision Tree, SVM, Naïve Bayes).
Clustering – Groups similar data points (K-Means, Hierarchical Clustering, DBSCAN).
Dimensionality Reduction – Reduces the number of features (PCA, t-SNE, LDA).
3️⃣ Model Training & Evaluation
Train-Test Split – Dividing data into training and testing sets.
Cross-Validation – Splitting data multiple times for better accuracy.
Metrics – Evaluating models with RMSE, Accuracy, Precision, Recall, F1-Score, ROC-AUC.
4️⃣ Feature Engineering
Handling missing data (mean imputation, dropna()).
Encoding categorical variables (One-Hot Encoding, Label Encoding).
Feature Scaling (Normalization, Standardization).
5️⃣ Overfitting & Underfitting
Overfitting – Model learns noise, performs well on training but poorly on test data.
Underfitting – Model is too simple and fails to capture patterns.
Solution: Regularization (L1, L2), Hyperparameter Tuning.
6️⃣ Ensemble Learning
Combining multiple models to improve performance.
Bagging (Random Forest)
Boosting (XGBoost, Gradient Boosting, AdaBoost)
7️⃣ Deep Learning Basics
Neural Networks (ANN, CNN, RNN).
Activation Functions (ReLU, Sigmoid, Tanh).
Backpropagation & Gradient Descent.
8️⃣ Model Deployment
Deploy models using Flask, FastAPI, or Streamlit.
Model versioning with MLflow.
Cloud deployment (AWS SageMaker, Google Vertex AI).
Data Science Resources
👇👇
https://whatsapp.com/channel/0029Va4QUHa6rsQjhITHK82y
Like for more 😄
❤5
⏳ Every month you postpone learning a new skill...
Someone else is building one.
Don't wait for the market to force you to adapt.
Get ahead with E&ICT Academy IIT Roorkee's AI & ML Program.
✅ 6 Months | Online | Open for all backgrounds
✅ Live sessions from IIT professors & industry mentors
✅ Placement support through Masai's network of 5000+ companies
🗓 Entrance Test: 26th July
🔗 https://tinyurl.com/DS-26Jul-008
Someone else is building one.
Don't wait for the market to force you to adapt.
Get ahead with E&ICT Academy IIT Roorkee's AI & ML Program.
✅ 6 Months | Online | Open for all backgrounds
✅ Live sessions from IIT professors & industry mentors
✅ Placement support through Masai's network of 5000+ companies
🗓 Entrance Test: 26th July
🔗 https://tinyurl.com/DS-26Jul-008
❤6
Machine Learning & Artificial Intelligence | Data Science Free Courses
⏳ Every month you postpone learning a new skill... Someone else is building one. Don't wait for the market to force you to adapt. Get ahead with E&ICT Academy IIT Roorkee's AI & ML Program. ✅ 6 Months | Online | Open for all backgrounds ✅ Live sessions…
Last 6 Hours Remaining!
Before the application closes for E&ICT IIT Roorkee AI & ML Program.
Don't miss out on the chance to:
• Learn live from IIT professors & industry experts
• Build real AI projects
• Get Placement Support from Masai.
Register NOW
Before the application closes for E&ICT IIT Roorkee AI & ML Program.
Don't miss out on the chance to:
• Learn live from IIT professors & industry experts
• Build real AI projects
• Get Placement Support from Masai.
Register NOW
❤5🎉1