🧠Nvidia releases OpenReasoning-Nemotron model

OpenReasoning-Nemotron-32B is a reasoning-focused LLM based on Qwen2.5-32B-Instruct. It has been further trained to generate solutions in mathematics, programming, and natural sciences. The model comes in 1.5B, 7B, 14B, and 32B parameter versions.

OpenReasoning-Nemotron is available for both commercial and research use.

💻Qwen3-Coder released — a powerful coding model from Alibaba

Qwen3-Coder-480B-A35B-Instruct is currently the most powerful open-source model for code generation. According to benchmarks, it performs at the level of Claude 4 Sonnet. It supports a context length of up to 1 million tokens.

For now, it’s available only as a free web version: https://chat.qwen.ai/

📜Google releases AI model for decoding ancient texts

The Aeneas model helps historians restore and decipher inscriptions from the Roman Empire. It analyzes text and photos of damaged tablets, finds similar inscriptions, and suggests what might have been written, when, and where.

The model draws on a database of 176,000 inscriptions, can reconstruct fragments of unknown length, and identifies the province and date with high accuracy. For example, it identified two likely timeframes for the creation of Augustus’ Res Gestae — both aligned with leading scholarly theories.