#Inference
Decision Models Benchmark in Russian: How Specialized AI Architectures Are Replacing Classical LLMs
A large-scale test of 12 decision models across 5,000 Russian-language tasks demonstrated the superiority of local open weights over cloud giants: accuracy reached 98.14%, and encoder inference latency dropped to 15 ms.
The Evolution of Attention in LLMs: How Engineers Scale Context to Millions of Tokens
We explore why the era of pure Multi-Head Attention is over and how a combination of MLA, learned sparsity, history compression, and FlashAttention 4 enables models to efficiently process ultra-long contexts.
NVIDIA and Over 100 Partners Launch Open Platform to Ensure Safe AI Agents
NVIDIA has introduced an open AI agent safety platform with more than 100 partners. The solution uses OpenShell runtime sandboxing and BlueField hardware monitoring to stop agents from escaping controls or deviating from intended actions.
Соавтор ChatGPT создал модель Jev: мгновенные типизированные решения вместо генерации текста
Стартап TypeSafe представил специализированную ИИ-модель Jev. Она заменяет медленные диалоговые LLM в программном коде на параллельную обработку данных с детерминированным результатом.
LLM Generation Speed on Apple M6, M5 Pro, Max, and Ultra: Detailed Performance Calculation
Analyzing the memory bandwidth of upcoming Apple Silicon chips allows for accurate advance estimates of local LLM performance — from lightweight Llama models to the massive 671-billion-parameter DeepSeek V3.
The Real Speed of Speculative Decoding: Why LLMs Don't Accelerate as Promised in Papers
The 6x inference speedup claimed in research papers often turns into an actual slowdown in practice. We explore how speculative decoding really works and what drives real-world performance gains.