AI in week 39.

CO

Claude Opus 5.5

Model · Anthropic

13

Anthropic's Claude Opus 5.5 model for agentic coding and knowledge work, priced lower per task than Opus 5.

GS

GPT-6 Sol

Model · OpenAI

4

OpenAI's mid-tier GPT-6 model, positioned below GPT-6 Astra and above Luna, priced at $2/$10 per million tokens.

S5

Step 5 Preview

Model · StepFun

2

StepFun's Step 5 Preview: a large MoE model (reported ~600B total, 27B active, 1M context) for coding and reasoning.

1B-parameter model from Paradigma specialized for hard competitive math problems, released with code and weights.

DN

Dots3 Note

Model · Rednote

1

Open MoE language model from Rednote (Xiaohongshu) with 208B total and 16B active parameters.

Open-weight cybersecurity model from Aikido, fine-tuned from GLM 5.3 and compressed to 328 GB.

QI

Qwen Image 2.1

Model · Alibaba

2

Alibaba Qwen's image generation and editing model, accessible through Qwen Chat and used as an open editing baseline.

MI

Ming Image 0.1 Design

Model · inclusionAI

1

Open-weights 6B text-to-image model for UI, infographics and posters, with 2K output and RGBA transparency.

Tiny 100M-parameter open-weights text-to-image diffusion transformer from SupraLabs, generating 256×256 images.

Experimental LoRA for Qwen-Image-2.1 that improves character consistency in anime-style image editing.

GO

Gemini Omni in Google Vids

App · Google

1

Google Vids' Gemini Omni feature generates and edits video clips with native audio and avatars inside Google Workspace.

Google Gemini 3.8 Live feature pairing real-time video generation with speech for conversational face-to-face avatars.

App that cuts long videos into multiple short clips from a typed text prompt, with one-click download.

R3

Ray 3.2

Model · Luma AI

1

Luma AI's Ray 3.2 video generation model with frame-level control and cross-cut continuity, via app and API.

G3

Gemini 3.8 Flash TTS

Model · Google

2

Google's Gemini 3.8 text-to-speech model for expressive, controllable speech generation via the Gemini API and AI Studio.

GL

GPT Live 1

API · OpenAI

1

OpenAI API for realtime voice assistants that keep talking while background agents work.

ST

Spotify Taste Profile

App · Spotify

1

Spotify feature that shows your inferred listening tastes and lets you adjust them by describing changes in text.

WO

WorldCrafter

Model · Tencent ARC Lab

2

Video world model from Tencent ARC Lab that turns an image or text into camera-controlled explorable scenes with 3D-aware memory.

GA

GAE

Model

1

Research autoencoder with a geometry-native latent space for 3D-consistent video and world generation.

MS

Mira Scene

Model

1

Model that reconstructs a single image into a 3D scene with separate, individually usable objects.

OP

OpenMuse

Agent

1

Open-source MIT-licensed personal agent app with browser, terminal, files and Gmail/Calendar, built on CopilotKit and AG-UI.

ZA

Zapier

App · Zapier

1

No-code automation platform connecting 9,000+ apps with deterministic workflows, AI steps and agents.

CM

Claude Managed Agents

Agent · Anthropic

1

Anthropic's hosted service for building and running Claude-based agents with custom harnesses in the cloud.

Meta Horizon tools that turn a text prompt into a 2D or 3D mobile game, via a phone app (Create) and a browser (Studio).

Tool that assembles web pages in real time from a preconfigured UI element library, as shown in creator demos.

YM

YouTube Made On YouTube AI features

App · YouTube

1

AI features announced at Made on YouTube: Ask YouTube search, custom feeds, Ask Music, auto dubbing and creator tools.

MC

Microsoft Copilot (new)

App · Microsoft

1

Microsoft's redesigned Copilot app combining chat, Cowork, code and autopilot modes across web, desktop and mobile.

Email prioritization feature that sorts an inbox by response priority, shown in a creator demo of Jev.

App that automatically removes ads, cookie banners and upsells from web pages, using your own API key.

Open-source fuzzy find-in-page tool that highlights semantically matching content rather than exact keywords.

F3

Flux 3 Action

Model · Black Forest Labs

1

7B open-weights world action model from Black Forest Labs that jointly predicts video and actions for robot control.

UD

Unitree Dex 5S

Hardware · Unitree

1

Human-sized dexterous robot hand from Unitree with 22 degrees of freedom and impact torque protection.

SA

Skild AI S1 soccer self-play

Research · Skild AI

1

Skild AI research result: humanoid soccer policy trained by self-play in Isaac Sim on the S1 model, transferred to a real robot.

Google research initiative to put TPU-equipped satellites in orbit as a prototype space-based AI data center.

RM

Ray-Ban Meta Audio

Hardware · Meta

1

Camera-free Ray-Ban Meta smart glasses with open-ear speakers, microphones and Meta AI voice assistant.

MR

Meta Ray-Ban Display

Hardware · Meta

1

Meta smart glasses with a monocular in-lens display, AI assistant and Meta Neural Band wristband controller.

Handheld keychain device from Meta for giving instructions to the Muse AI agent; announced, still in production.

Meta's announced lightweight VR glasses with a tethered compute and battery puck, reported for spring 2027.

CE

Cerebras

Hardware · Cerebras

1

Wafer-scale AI chips (CS-series) and a cloud inference API serving open models at high token speeds.

GR

Groq

Hardware · Groq

1

Inference provider built on its custom LPU chips, offering fast LLM inference via cloud API (GroqCloud) and dedicated capacity.

CG

Claude gene discovery (novel enzyme system)

Research · Anthropic

1

Anthropic research result: Claude agents found a CRISPR-like array-associated reverse transcriptase system in phage DNA.

CV

CLM v0.1 8B

Model

1

Open-weights contrastive "System One" scorer: projection heads on a frozen Qwen3-8B that rank actions against states.

Open multimodal embedding models from AIDC-AI that map text, images, video and documents into a shared vector space.

C3

ClusterMAX 3.0

Research · SemiAnalysis

1

SemiAnalysis's rating system that ranks GPU cloud (neocloud) providers, with CoreWeave and Nebius highlighted.

Anthropic tooling to check whether Claude skills and plugins are up to date with the newest model and update them.

Community GGUF quantizations of Qwen-Image-2.1 for low-VRAM text-to-image and image editing in ComfyUI.

App that suggests the emojis most associated with whatever text you type, shown in a YouTube demo of Jev use cases.

Demo app that generates color palettes from any typed concept, such as 80s disco, Mario or the Matrix.