AI in week 21.

G3

Gemini 3.5 Flash

Model · Google

5

Google's faster, lower-cost Flash model in the Gemini 3.5 family, aimed at agentic and coding workloads.

Q3

Qwen 3.7 Max

Model · Alibaba

2

Alibaba's Qwen flagship large language model, offered as a hosted proprietary model via Qwen Chat and Alibaba Cloud.

GE

Gemma

Model · Google

2

Google DeepMind's family of open-weights models built from Gemini technology, from phone-sized to laptop-class variants.

HY

HY-MT2

Model · Tencent

1

Tencent's open multilingual translation model family (1.8B, 7B, 30B-A3B MoE) covering 33 languages.

2B-parameter open-weights video-language model that writes scene descriptions and timestamped event captions.

L2

L2P

Model

1

Pixel-space text-to-image diffusion model transferred from latent diffusion models, without a VAE, supporting 4K output.

K2

Krea 2 LoRAs

App · Krea AI

1

LoRA training on Krea's platform to teach Krea 2 a custom style, character, face or product from a few images.

Google image creation and editing app built on Nano Banana, with object-level edits and Docs, Slides and Drive integration.

GO

Gemini Omni

Model · Google

6

Google's Omni video generation model, available in the Gemini app and used for video creation and editing in Google Vids.

CO

CogOmniControl

Research

1

Research framework for controllable video generation from sketches, pose or line art, a reference image and a text prompt.

LA

Lance

Model · ByteDance

1

ByteDance research model unifying image and video understanding, generation and editing in a 3B active-parameter network.

FA

FashionChameleon

Model · Alibaba

1

Research model for real-time video virtual try-on that swaps garments mid-generation on a single GPU.

LV

LongCat Video Avatar 1.5

Model · Meituan

1

Open-weights audio-driven avatar video model from Meituan, built on LongCat-Video, with 8-step distilled inference.

MA

Mega ASR

Model

1

Speech recognition model trained for noisy, reverberant and distorted real-world audio, with RL post-training.

Meta research model that generates audio directly in raw waveform space, without a VAE, for video-to-audio synthesis.

Q3

Qwen 3.5 LiveTranslate

Model · Alibaba

1

Alibaba Qwen model for live speech translation across 60 languages that uses visual context and streams translations early.

Alexa+ feature that generates on-demand, podcast-style audio episodes on a chosen topic.

SP

Spotify Personal Podcasts

App · Spotify

1

Spotify feature for generating private AI-made audio episodes from a prompt and asking questions about podcasts.

SA

Stable Audio 3

Model · Stability AI

2

Stability AI's family of text-to-audio models for music and SFX, with open-weight Small and Medium variants and an API-only Large.

SA

Spotify AI Covers and Remixes

App · Spotify

1

Spotify's licensed feature for fan-made AI covers and remixes, developed with Universal Music Group and sharing revenue with artists.

AT

Atlas

Model · World Labs

3

World Labs' world model that generates and reconstructs explorable 3D worlds from images, text, video or 3D input.

LI

LiTo

Model · Apple

1

Apple research model that generates 3D assets with view-dependent appearance from a single input image.

PA

PanoWorld

Model

1

Research generative world model that creates consistent whole-house panorama tours from a floorplan and style reference.

RE

ReactiveGWM

Model

1

Research game world model that assigns player or NPC roles per character and generates reactive NPC behavior from rules.

PG

Project Genie

App · Google DeepMind

1

Google DeepMind app for generating interactive, explorable worlds from prompts, including places grounded in Street View.

OP

OpenClaw

Agent

5

Open-source personal AI assistant that runs on your own machine and connects to chat apps like WhatsApp, Telegram and Slack.

AC

AI co-scientist

Agent · Google DeepMind

2

Gemini-based multi-agent system from Google that generates, debates and ranks scientific hypotheses for researchers.

Google shopping cart spanning multiple merchants, with agent checkout via the Agent Payments Protocol (AP2).

C2

Composer 2.5

Model · Cursor

4

Cursor's in-house coding model, used in the Cursor editor and by sub-agents for lower-cost coding tasks.

AN

Antigravity

App · Google

3

Google's agent-first development platform with desktop app, terminal CLI, IDE and SDK for running coding agents.

GA

Google Antigravity 2.0

App · Google

2

Google's agent-first development platform with desktop app, CLI, IDE and SDK for running parallel coding agents.

ST

Stitch

App · Google

1

Google's AI design tool that generates mobile and web app UIs from text prompts.

Google DeepMind AI agent that finds software vulnerabilities and generates, tests and verifies security patches.

GA

Google AI Mode connected apps

App · Google

3

AI Mode in Google Search can connect to third-party apps such as Instacart to act on a prompt, e.g. a grocery list.

GV

Gemini voice in Gmail, Docs and Keep

App · Google

2

Conversational Gemini voice interaction inside Gmail, Google Docs and Google Keep for paid Google AI plan subscribers.

PO

Pomelli

App · Google

1

Google Labs tool that builds a brand profile from a website and generates on-brand marketing content for small businesses.

GB

Granola Briefs

App · Granola

1

Granola feature that prepares a pre-meeting Brief for external meetings from calendar data and past conversations.

AI

AI Inbox

App · Google

1

Gemini-powered inbox assistant in Gmail that summarizes threads, drafts replies and answers questions about your email.

GD

Gemini Daily Brief

App · Google

1

Personalized morning digest in the Gemini app that summarizes Gmail, Calendar and other connected apps.

UG

Unitree G1 voice control

Hardware · Unitree Robotics

1

Unitree G1 is a compact humanoid robot from Unitree Robotics, shown in demos following spoken commands.

Open-source 3D-printable humanoid robot platform from the Hugging Face LeRobot project, with parts list and software.

GG

Google Gemini AI Glasses

Hardware · Google

2

Gemini-powered Android XR smart glasses from Google, built with Samsung, Qualcomm, Gentle Monster and Warby Parker.

T8

TPU 8t

Hardware · Google

1

Google's eighth-generation TPU for large-scale AI pre-training, scaling to 9,600 chips per superpod.

T8

TPU 8i

Hardware · Google

1

Google's eighth-generation Cloud TPU for post-training and low-latency inference of large MoE models.

Open-source DNA foundation model family from Hugging Face, 500M to 8B parameters, with ~400k base-pair context.

FG

Flash GRPO

Research

1

Single-step GRPO training framework for aligning video diffusion models with human preferences, tested on Wan2.1 1.3B–14B.

Google suite of science tools whose "science skills" connect Gemini to 30+ life science databases such as UniProt and AlphaFold.

GA

Google AI watermark removal

App · Google

3

Google Gemini feature reported to let users remove visible watermarks from AI-generated media while SynthID stays embedded.