DeepSeek‑R1 Shows RL‑Only Reasoning Matches o1 on AIME

Google’s latest Gemini API updates bring faster real-time voice conversations, native audio-to-audio interaction, function calling, and new text-to-speech models—giving developers more practical tools for building voice-first AI products.

LFM2.5-VL-DSpark accelerates vision-language models with a novel architecture and optimized implementation, outperforming state-of-the-art models in various benchmarks. This breakthrough is made possible by the integration of LiquidAI's LFM2.5-VL architecture with DSpark, a distributed computing framework.

This article presents a novel approach to improving OCR faithfulness by leveraging gated and attenuated on-policy distillation. The proposed method demonstrates significant improvements in OCR accuracy and faithfulness on various benchmark datasets.