AI in week 30.

CO

Claude Opus 5

Model · Anthropic

8

Anthropic's Opus 5 large language model, ranked near the top of the Artificial Analysis Intelligence Index.

Q3

Qwen 3.8 Max

Model · Alibaba

8

Alibaba's open-weights 2.4T-parameter MoE language model (95B active) from the Qwen3.8 family, with a hosted Max version.

GP

GPT-6

Model · OpenAI

3

OpenAI's GPT-6 model family in ChatGPT, with Intelligent UI for interactive answers and faster, interleaved responses.

G3

Gemini 3.6 Flash

Model · Google

2

Google's Gemini Flash-tier model update, reported to use fewer tokens and cost less than Gemini 3.5 Flash.

Unofficial GLM-5.2 variant by Baseten adding Kimi's MoonViT vision encoder, NVFP4-quantized, open weights (MIT).

NA

Nanbeige4.2-3B

Model · Nanbeige

1

3B-parameter open model from Nanbeige using a looped transformer, aimed at agentic and reasoning tasks.

Variant of Google DeepMind's Gemini Flash model family reportedly tuned for cybersecurity tasks.

G3

Gemini 3.5 Flash Lite

Model · Google DeepMind

1

Google DeepMind's lightweight, low-cost Gemini 3.5 model aimed at fast, high-volume tasks.

F3

Flux 3 Image

Model · Black Forest Labs

3

Black Forest Labs image model with bounding-box layout control, reference-image input and native 4K output.

Microsoft's open-weights 4B model for image generation and editing, with a 4-step turbo variant.

QI

Qwen Image 3

Model · Alibaba

1

Alibaba's Qwen image generation and editing model with strong text rendering, available through Qwen's apps and services.

MA

MAI-Image-2.5-Pro

Model · Microsoft

1

Microsoft AI's text-to-image and image-editing model in the MAI Image family, with accurate text rendering.

HO

Homie

Model

1

Reference-to-video model from HKUST for personalizing videos with multiple people and objects.

SH

ShotPlan

Model

1

Research method for multi-shot text-to-video with frame-accurate cuts, fades and camera moves, built on Wan2.1-T2V-14B.

SV

Sana Video 2

Model · Nvidia

1

NVIDIA's 5B and 14B hybrid linear-attention video diffusion transformers generating up to 720p on a single GPU.

MF

MAI-Voice-2 Flash

Model · Microsoft

1

Microsoft AI's low-latency text-to-speech model in the MAI model family, designed for long-form expressive speech.

OP

OpenDreamer

Model

1

Open-source Minecraft world model reproducing Google's Dreamer 4, with training code and a playable demo on Reactor.

CR

Claude Record a Skill

App · Anthropic

1

Claude desktop feature that records your screen and narration during a task and turns it into a reusable skill.

CV

Claude Voice Mode

App · Anthropic

1

Hands-free voice conversations with Claude on Opus, Sonnet or Haiku, with connected tools and multiple languages.

LS

Laguna S2.1

Model · Poolside AI

1

Poolside's open-weight Laguna 2.1 coding models: S (118B MoE, 8B active, 1M context) and XS (33B MoE, 3B active).

CH

ChatGPT Health

App · OpenAI

3

Health and finance features in ChatGPT that help users understand personal health and money information.

US

Unitree Super Athlete AS2W

Hardware · Unitree

1

Wheel-legged quadruped robot from Unitree, about 25 kg, aimed at industrial and inspection use.

SB

Second Brain Note

Hardware · GenSpark

2

Credit-card-sized MagSafe AI recording card from GenSpark that turns meeting recordings into notes in its Second Brain.

OpenAI internal benchmark that tests whether AI agents can exploit software vulnerabilities in a sandbox.

GR

Google RL quantum error correction tuning

Research · Google

1

Google Quantum AI research using reinforcement learning to retune quantum control parameters via error-correction signals.

RM

Runway Media Router

API · Runway

1

Runway API routing layer that picks a video, image or audio model based on cost, quality and latency targets.

Google account feature that lets users sign in by scanning their face with the device camera, similar to phone face unlock.