Insight tag

multimodal models

5 related stories.

Stories

Original explanatory diagram

News analysisRelease: January 27, 2026Published: October 24, 20263 min read

Kimi K2.5 combined native multimodality with a research-preview agent swarm

Moonshot AI introduced Kimi K2.5 on January 27, 2026, pairing vision-text pretraining, open weights, and a beta agent-swarm mode with dynamic subagents.

Read analysis

Original explanatory diagram

News analysisRelease: December 16, 2025Published: October 24, 20263 min read

Meta AI SAM Audio audio AI multimodal models

Meta introduced SAM Audio with text, visual and temporal prompts for source separation

Meta introduced SAM Audio on December 16, 2025, pairing multimodal audio separation with PE-AV, an evaluation benchmark and a reference-free judge model.

Read analysis

Original explanatory diagram

News analysisRelease: May 20, 2025Published: October 23, 20263 min read

Google Gemini 2.5 Flash thinking budgets multimodal models

Gemini 2.5 Flash update made thinking controls a product variable

Google described an updated Gemini 2.5 Flash Preview at I/O on May 20, 2025, pairing speed-oriented positioning with thought summaries and developer controls.

Read analysis

Original explanatory diagram

News analysisRelease: April 5, 2025Published: October 23, 20263 min read

Meta Llama 4 open weights multimodal models

Llama 4 showed why open weights still need deployment boundaries

Meta released Llama 4 Scout and Maverick in April 2025. Their multimodal, mixture-of-experts design made model cards, licensing and workload conditions part of adoption evidence.

Read analysis

Original explanatory diagram

News analysisRelease: March 27, 2025Published: October 23, 20263 min read

Qwen Qwen2.5-Omni multimodal models speech generation

Qwen2.5-Omni put multimodal streaming on a joined input-output path

Qwen released Qwen2.5-Omni on March 27, 2025, describing one model that processes text, images, audio and video while streaming text and speech.

Read analysis