AI News

An Inc.com headline says Anthropic banned cruelty toward Claude, but the policy, its scope and the company’s reason are not confirmed by available details.

A developer’s account says DeepSeek 4.1 Flash handles demanding coding work at low cost, but independent benchmarks and industry reaction remain unclear.

A South China Morning Post headline reports Chinese-language options settings in Claude, but does not specify their scope, availability or release date.

The Register reports a study favored Claude over OpenAI on subscription value, but the study, methods and plans compared are not identified.

OpenAI released 722 manuscripts across 372 result families. Their claims remain unverified, and mathematicians must determine what is correct and useful.

Falcon-Emirati-7B adapts Falcon-H1-Arabic with Emirati dialect text, cultural material and synthetic examples; independent results were not provided.

Mistral Large 4 Preview scores 38 on Artificial Analysis’s Intelligence Index, below leading US and several Chinese models, as weights remain unreleased.

Mistral Large 4 scores 38.4 on Artificial Analysis, but trails leading US and Chinese models and raises cost, verbosity and reliability concerns.

A Times of India headline reports Claude inference in India through AWS, but availability, regions and data-handling terms remain unconfirmed.

Reflection says Beam combines 501 billion total parameters with large-scale reinforcement learning, targeting coding and agentic work. Weights are due later this month.