AI in week 23.

G4

Gemini 4 Carbon

Model · Google

14

Google DeepMind's Gemini 4 Carbon, a frontier Gemini 4 language model with a staged rollout to paid users and API.

MM

MiniMax M3

Model · MiniMax

4

Open-weights native multimodal MoE model from MiniMax, ~428B parameters (~23B active) with 1M-token context.

CO

Claude Opus 4.5

Model · Anthropic

4

Anthropic's Claude Opus 4.5 model for agentic coding, tool use and knowledge work, available via app and API.

CO

Claude Opus 4.6

Model · Anthropic

3

Anthropic's Claude Opus 4.6 large language model for coding, agentic work and knowledge tasks, via Claude apps and API.

CD

ChatGPT Dreaming

App · OpenAI

2

OpenAI's ChatGPT memory system that synthesizes context from past chats in the background to keep memory current.

G4

Gemma 4 12B

Model · Google

2

Google's open-weights 12B encoder-free multimodal model in the Gemma 4 family, taking text, image and audio input.

N3

Nemotron 3 Ultra

Model · NVIDIA

2

NVIDIA's open-weight ~550B mixture-of-experts LLM with a hybrid Mamba-transformer design, aimed at agentic workloads.

MT

MAI Thinking 1

Model · Microsoft

1

Microsoft AI's mixture-of-experts reasoning model, in private preview via Microsoft Foundry.

Q3

Qwen 3.7 Plus

Model · Alibaba

1

Alibaba's proprietary multimodal agentic model in the Qwen 3.7 series, aimed at long-horizon coding and tool-use tasks.

MM

Microsoft MAI Models (7 new)

Model · Microsoft

1

Microsoft AI's in-house MAI model family covering reasoning, code, speech, transcription, image and cybersecurity.

MC

Mayo Clinic Microsoft Healthcare AI Model

Model · Microsoft

1

Frontier healthcare AI model co-developed by Mayo Clinic and Microsoft, owned by Mayo, planned for Azure Foundry APIs.

DE

DeepSeek

Model · DeepSeek

1

Chinese AI lab behind the open-weight DeepSeek-V4 and DeepSeek-R1 models, with a chat app and low-cost API.

Google's 2B-parameter Gemma 2 instruction-tuned model, repackaged as a text encoder for PixelDiT and PiD workflows in ComfyUI.

ZI

Z-Image

Model

2

Open-weights 6B text-to-image model family from Alibaba's Tongyi-MAI lab, with a fast 8-step Turbo variant.

SL

Stable Layers

Research · Stability AI

1

Stability AI research method that fine-tunes Qwen-Image-Layered for image layer decomposition using VLM-scored RL.

R2

Reve 2

Model · Reve

1

Closed-source text-to-image and editing model from Reve that plans composition with structured bounding boxes.

K2

Krea 2 Turbo

Model · Krea

1

Faster variant of Krea's Krea 2 image model, reported to generate images in about 2 seconds.

I4

Ideogram 4.0

Model · Ideogram

1

Ideogram's open image model with text rendering and transparent backgrounds, available via the Ideogram app and API.

SE

SeedVR2

Model · ByteDance

1

ByteDance's open-weights one-step diffusion transformer for restoring and upscaling video and images.

NA

NAVA

Model · Baidu

1

Baidu ERNIE team's 6.3B joint audio-video generation model with native audio-visual alignment and timbre-controlled speech.

BE

Bernini

Model · ByteDance

1

ByteDance unified video generation and editing framework pairing an MLLM semantic planner with a DiT renderer.

ST

StreamChar

Model · Alibaba

1

Alibaba research model for real-time, long-horizon streaming character audio-video generation.

RA

Runway Aleph 2.0

Model · Runway

1

Runway's video-to-video model for editing and transforming existing footage with text prompts.

GI

Grok Imagine 1.5

Model · xAI

1

xAI's Grok Imagine video generation model with dialogue support, offered through Grok plans.

EV

ElevenLabs V4

Model · ElevenLabs

6

ElevenLabs' expressive text-to-speech model with inline audio tags, 90+ languages and Professional Voice Clone support.

WA

WavTTS

Model · ByteDance

1

Research zero-shot TTS model that generates speech by modeling raw waveforms directly with flow-matching DiT.

HA

Higgs Audio v3

Model · Boson AI

1

Text-to-speech model from Boson AI with inline tags for emotion, style, speed, pitch, pauses and sound effects.

Open-source text-to-speech voice model described by creators as highly emotive and human-sounding.

MR

Magenta RealTime 2

Model · Google

1

Open-weights 2.4B live music model from Google controllable via MIDI, audio and text, running locally on Apple Silicon.

MA

Mamma

Model

1

Markerless multi-person motion capture that recovers SMPL-X body, hands and shape from multi-view video.

OM

OmniDreams

Model · NVIDIA

1

NVIDIA's real-time generative world model for closed-loop autonomous vehicle simulation, built on Cosmos.

PA

PaGeR

Model · Google

1

Predicts depth, surface normals and sky masks from a single 360° panorama using a cubemap-adapted 3D foundation model.

DV

Deja View

Model · NVIDIA

1

NVIDIA's 117M-parameter looped transformer for multi-view 3D reconstruction of camera poses and dense geometry.

HA

Hermes Agent

Agent · Nous Research

4

Open-source, self-hostable AI agent from Nous Research with persistent memory, skills, scheduling and multi-platform messaging.

O2

OpenClaw 2.0

Agent · OpenClaw

4

Open-source personal AI agent that runs on your own machine and is controlled from chat apps like WhatsApp and Telegram.

Microsoft's always-on personal AI agent for Microsoft 365 apps, built on OpenClaw technology; early testers only.

Open-source framework for orchestrating teams of AI agents to run an autonomous, company-like workflow.

Startup product pitched as an AI system that builds and runs a company autonomously with minimal human input.

CC

Claude Code

App · Anthropic

18

Anthropic's Claude assistant and agentic coding tool, with chat apps, Projects, skills, connectors and Claude Code.

DH

DeepSeek harness

App · DeepSeek

3

DeepSeek's desktop agentic harness for running the DeepSeek models on coding projects, reportedly several in parallel.

GC

GitHub Copilot App

App · GitHub

1

Desktop app from GitHub for running and supervising parallel Copilot coding agents across connected repositories.

CW

Codex Windows Computer Use & Remote Control

App · OpenAI

1

Codex app update adding computer use on Windows and remote control of Windows devices from the ChatGPT mobile app.

Microsoft's coding-focused language model, positioned as a faster, more token-efficient alternative to Claude Haiku 4.5.

CP

Codex Plugins and Sites

App · OpenAI

1

OpenAI Codex role-specific plugins, shareable Sites and annotations for non-developer knowledge work.

UH

UBTech humanoid companion robot

Hardware · UBTech

1

UBTech's humanoid robot line for commercial, education and home use, including a teased companion humanoid.

Industrial-grade, all-weather humanoid robot from Deep Robotics with dust and water protection.

C3

Cosmos 3

Model · NVIDIA

1

NVIDIA's open omni-model family for physical AI: world simulation, reasoning and action generation for robots and AVs.

NR

NVIDIA RTX Spark

Hardware · NVIDIA

4

NVIDIA superchip platform for Windows laptops and small desktops, pairing a Grace CPU with a Blackwell RTX GPU.

M2

Majorana 2

Hardware · Microsoft

1

Microsoft's second-generation topological quantum chip with a lead-based materials stack, developed with Microsoft Discovery.

PS

Project Solara

Hardware · Microsoft

1

Microsoft platform and early hardware concepts for agent-first devices, aimed initially at enterprise.

WA

When AI Builds Itself (Anthropic recursive self-improvement paper)

Research · Anthropic

1

Anthropic essay on delegating AI development work to AI systems and the risks of recursive self-improvement.

CO

CORE-Bench

Research

1

Benchmark of 270 tasks testing whether AI agents can computationally reproduce results from scientific papers.

CO

ComfyUI

App · Comfy Org

7

Node-based visual AI engine for running image, video, audio and 3D generation workflows locally, in the cloud or via API.