ElevenLabs has released Scribe v3 and Scribe v3 Live, two speech recognition models designed to transcribe accurately through background noise and crosstalk. Scribe v3 Live uses conversational context to help voice agents avoid interrupting callers mid-sentence. According to benchmarks from Artificial Analysis, Scribe v3 narrowly beats competitors including OpenAI, Google, and Microsoft on accuracy metrics (The Decoder).
Google Deepmind introduced the Gemini Decisions API, powered by a specialized model called Gemini Decision 1 that evaluates up to 1,000 closed-ended questions in a single pass across text, images, audio, and video. The model returns results in 150 to 250 milliseconds at approximately five times lower cost than Gemini 3.5 Flash-Lite, priced at $0.10 per 1 million input tokens (The Decoder, MarkTechPost). Separately, Google Deepmind released AlphaGenome 2, a genomics model trained on more than 107 million cell profiles that analyzes genetic variants at single-cell resolution, identifying nearly twice as many disease-relevant gene changes as its predecessor and available via API for non-commercial research (The Decoder).