AI — Page 16

The new model, now on all paid Claude plans, hits state‑of‑the‑art scores on coding, ARC‑AGI‑3 and alignment tests, according to Anthropic.

In a July 24 X post, AMD disclosed a new MI455X GPU that packs 432GB of HBM4, enough to store FP8 weights of a trillion‑parameter model in a single server.

The open‑source community unveiled LLaDA2.2-flash on July 24th, combining mixture‑of‑experts diffusion with token‑level edit operations for on‑the‑fly correction.

The open-source agent platform lets developers continue a halted coding session when hitting usage caps on OpenAI's Codex or Anthropic's Claude Code.

Nvidia's chief executive joined Satya Nadella and other industry leaders in signing an effort linking accessible model weights to US competitiveness, economic opportunity, and national security.

The spring deal gives Midjourney an established mobile audience as David Holz expands beyond image models, Discord and the web.

The Zurich startup added factory data and an action decoder to Black Forest Labs' multimodal backbone for Audi manipulation tests.

The new format cuts token usage by up to fivefold and lets agents learn across Claude Code, Codex and Letta Code sessions.

Enze Xie announced the model on X, highlighting a hybrid attention backbone and a new training pipeline that cuts compute while preserving quality.

The update adds unspecified upgrades and follows strong user adoption of Sakana AI's language models.

Velmeni is extending from 2D X-rays into CBCT, betting a unified imaging, voice and claims platform can beat standalone tools.

Artem Kukharenko posted on X that Palabra.ai's new text-to-speech model hit 104 ms time to first audio, half the industry average.

The video‑editing platform announced the new tool on July 23, 2026, and is giving the first 100 users 500 free AI credits.

One model brought flash; the other brought control. Across prompt-following tests that punish sloppiness—text, counting, layout, object assignment, and emotional specificity—Rundiffusion Photo Flux separated itself decisively.

The new hybrid‑reasoning model runs with 5.1B active parameters per token and is free on OpenRouter through Aug 3, 2026.

Ruikai Peng posted a preview of Rolling in the Diffs, an LLM‑based vulnerability‑discovery system he says uncovered over 60 flaws while still in high school.

The new voice option lets Claude run longer, tool‑integrated conversations and pick the latest text model by default.

The new streamTranscribe SDK method lets developers feed live audio and receive instant transcripts, expanding Vercel's AI edge capabilities.

The startup's Sound Effects 1.0 claims top performance on multiple video-to-audio benchmarks and demos for phone footage, ads and games.

The company reports the Gemini consumer AI app now serves almost a billion users, with daily usage three times higher than a year ago.

On July 23 2026, Dev released a benchmark that measures how well large language models can operate AutoCAD via computer-use.

The new models deliver higher‑fidelity image generation and faster, cheaper speech synthesis for Microsoft products.

Google's first ATLAS report, based on 15 million de‑identified interactions, offers the most comprehensive look at how users adopt its AI tools.

The U.S. rollout lets adults connect Apple Health and supported medical records, with permission controls and a promise that connected data will not train foundation models.