OpenAI
GPT 5
OpenAI has released their new foundation model, GPT 5. It is a unified (text-image) model and OpenAI claims it is especially good at coding. One important statistics reported by OpenAI is the hallucination rate, which seems to have gone below 1%.
GPT-OSS
OpenAI has released a fully open-weight model, GPT-OSS, for the first time. It comes in two variants, one with 120 Billion parameters and the other with 20 Billion parameters.
Google
Imagen 4 Fast
Google has released a “Fast” version of their Imagen image-generation software, which is more cost-effective.
Meta
Dino v. 3
Meta has put their Dino model open-source. This model excels in various image processing tasks (object detection, image segmentation, and depth estimation).
Anthropic
Claude Opus 4.1
Claude now entertains a one-million token context window, thus enabling it to use a much longer and useful context in complex tasks like coding.
DeepSeek
DeepSeek v. 3.1
DeepSeek has released their new version of the DeepSeek model. It is a hybrid reasoning model (only using reasoning for difficult tasks as needed).