At WWDC25, Apple announced a framework that lets apps use the on-device Apple Intelligence model. Availability depended on new operating-system releases and compatible hardware.
At I/O 2025, Google unveiled Veo 3, which it calls the first video model with native audio, Imagen 4 and AI Ultra; the existing AI Premium plan became Google AI Pro.
Meta introduces Llama 4 Scout and Maverick, its first natively multimodal models built with a mixture-of-experts architecture. The launch comes amid questions about Maverick’s LMArena score.
Google introduced gemini 2.5 pro experimental as the first model in its new family. The source bounds the event: arena rankings and vendor benchmarks are protocol snapshots, not a universal hierarchy. The story teaches how to compare models with version, date, dataset, judge, tools and cost fixed.
Openai integrated native image generation into gpt-4o inside chatgpt. The source bounds the event: selected examples show capability, not success rate; the vendor's own card lists limitations and controls. The story teaches how to test text, editing, identity, provenance and failures before using an image.
Google unveils Gemma 3, open models from 1B to 27B parameters. The 4B, 12B and 27B variants accept images and have 128k context; Google highlights single-accelerator performance.
OpenAI releases GPT-4.5, its largest and most knowledgeable model yet. It promises fewer hallucinations, while its tests show strengths distinct from reasoning models.
Black Forest Labs, founded by researchers behind Stable Diffusion technologies, launched FLUX.1 on August 1, 2024. Pro, dev and schnell offered different quality, cost and licensing trade-offs; choosing among them requires separating model, output and service.
Google brought Gemini 1.5 Flash to the free app on July 25. The Pro price cut arrived in September, not July.
AlphaProof and AlphaGeometry 2 solved four of the six problems at the 2024 International Mathematical Olympiad. Their 28 points amounted to a silver medal, just one point short of the gold cutoff.
On July 23, 2024, Meta released Llama 3.1 in 8-billion, 70-billion and 405-billion-parameter sizes. Downloadable weights increase control, but do not remove inference costs, the custom licence or the need to evaluate the whole system.
Google’s AI Overviews have recommended putting glue on pizza and eating rocks. The failures, which emerged after the US launch, expose the risk of turning text found online into authoritative-sounding answers.
This website uses cookies to improve the browsing experience. Cookie policy.
No results. Try different words.
