AI in week 22.

GP

GPT-5.5

Model · OpenAI

8

OpenAI language model available through ChatGPT, Codex and the API, used for coding, review and general tasks.

CO

Claude Opus 4.8

Model · Anthropic

7

Anthropic's Claude Opus 4.8 large language model, offered via Claude subscriptions and the Anthropic API.

CS

Claude Sonnet 4.6

Model · Anthropic

3

Anthropic's mid-tier Claude model, used as a general-purpose workhorse for coding, agents and knowledge work.

CO

Claude Opus 4.7

Model · Anthropic

3

Anthropic's Opus-tier Claude model for agentic coding and long-running knowledge work, offered via API and Claude apps.

GP

GPT-5.4

Model · OpenAI

2

OpenAI frontier language model used for coding, reasoning and agentic tasks, available via ChatGPT and the API.

S3

Step 3.7 Flash

Model · StepFun

1

Open-weights multimodal flash model from StepFun for agentic tasks involving images, interfaces, web search and long runs.

M1

MiniCPM5 1B

Model · OpenBMB

1

1B-parameter dense open-weights language model from OpenBMB for on-device use, with hybrid thinking and 128K context.

CH

Claude Haiku 4.5

Model · Anthropic

1

Anthropic's small, fast Claude model for low-latency, cost-sensitive text and coding workloads, with image input.

KK

Kimi K2.5

Model · Moonshot AI

1

Moonshot AI's open-weights 1T-parameter MoE multimodal agentic model with 256K context, vision input and agent swarm.

MI

MAI Image 2.5

Model · Microsoft

3

Microsoft AI's text-to-image and image-editing model, with a lower-cost Flash variant, offered via Microsoft Foundry.

PI

PiD

Model · Nvidia

2

NVIDIA pixel-diffusion decoder that turns latents into 2K–4K images, replacing VAE decode plus a separate upscaler.

CO

ControlLight

Model

1

Research model for low-light image enhancement with continuous, controllable brightness, built on Flux 2 Klein.

SE

Sega

Model

1

Training-free method that rescales attention across RoPE components to generate ultra-high-resolution images with diffusion transformers.

PI

PixlRelight

Model

1

Feed-forward single-image relighting model with physically based light control, using intrinsic conditioning from a coarse 3D scene.

PrismML's 1-bit/ternary compressed Flux 2 Klein text-to-image model (~1 GB) for offline on-phone generation.

Code-driven video rendering and motion-graphics library used with Claude to produce overlays, title cards and B-roll.

SC

Scope

Library · Tencent

2

Positional encoding method that gives video diffusion transformers precise camera control from an image and camera path.

P3

Pantheon 360

Model

1

3D-aware 360° video diffusion model that generates panoramic videos along user-defined camera paths from sparse panoramas.

IN

InstructAV2AV

Model

1

Research model for editing video and audio jointly from a text instruction, e.g. changing speech with lip sync or voice.

ED

ElevenLabs Dubbing v2

App · ElevenLabs

1

ElevenLabs' AI dubbing model that translates video and audio while carrying over the original speaker's voice and emotion.

SA

Spotify AI Persona Badge

App · Spotify

2

Spotify artist-profile badge and labeling for AI personas, based on artist self-disclosure.

EM

ElevenLabs Music v2

Model · ElevenLabs

1

ElevenLabs' text-to-music model for generating songs with vocals, inpainting and genre transitions, cleared for commercial use.

GW

Gamma World

Model · Nvidia

1

NVIDIA research model that generates real-time, action-controlled multi-agent video of shared worlds, from two to four players.

TR

TriSplat

Model

1

Feed-forward 3D scene reconstruction model that outputs triangle primitives directly usable as meshes for simulation.

GE

GenRecon

Research

1

Research method that reconstructs PBR-ready indoor 3D scene meshes from multi-view images using a Trellis.2 generative prior.

CU

CubePart

Model

1

Roblox research model that generates text-to-3D objects as separate part meshes from a user-defined part list.

PO

PhysX Omni

Model

1

Research framework generating simulation-ready 3D assets (rigid, deformable, articulated) with physical properties.

RH

Relightable Hollywood-ported Characters

Research · Google

1

Research method for free-view rendering and relighting of full-body dynamic humans from four sparse RGB views.

LI

Leonardo Image to 3D

App · Leonardo AI

1

Leonardo.Ai feature that generates 3D models from a single image, using multi-angle reference views.

PC

Perplexity Computer (local on Mac)

Agent · Perplexity

4

Perplexity's agent that can run locally on a Mac as well as on Perplexity's cloud computers.

AU

AutoScientists

Agent

1

Decentralized multi-agent LLM framework where agents self-organize into teams for long-running computational research.

RO

Robin

Agent · FutureHouse

1

FutureHouse multi-agent system that automates hypothesis generation and data analysis for experimental biology.

DW

Dynamic Workflows

App · Anthropic

2

Claude Code feature where Claude writes orchestration scripts that run tens to hundreds of parallel subagents with verification.

M3

Microsoft 365 Copilot redesign

App · Microsoft

1

Microsoft's AI assistant built into Word, Excel, PowerPoint, Outlook and Teams, grounded in work data via Work IQ.

Transformable consumer robot that switches between quadruped and bipedal humanoid modes, reportedly priced near 20,000 yuan.

AZ

Athena Zero

Hardware · RAI Institute

1

Humanoid robot platform from the RAI Institute, shown juggling and switching styles after brief real-world learning.

Self-service robot barber kiosks in Chinese cities that 3D-scan the head and cut hair automatically.

MJ

Meng Jiao Yi AI Pet Translator

Hardware · Meng Jiao Yi

1

Neck-worn AI device from Chinese startup Meng Xiaoyi that claims to translate pet sounds and behavior into human speech.

VM

vidIQ MCP

API · vidIQ

15

vidIQ's official Claude connector (MCP) for YouTube analytics, keyword research, trends and outlier detection.

DE

DeepSWE

Research · Datacurve

4

Benchmark of frontier coding agents on original, long-horizon software engineering tasks, by Datacurve.

SP

SWE-bench Pro

Research

2

Benchmark from Scale AI testing AI agents on long-horizon software engineering tasks drawn from real codebases.

BB

BES (Bidirectional Evolutionary Search)

Research

1

Search framework from Harvard/MIT coupling forward evolutionary candidate search with backward goal decomposition.

LO

LocateAnything

Model · Nvidia

1

Nvidia vision-language model for object detection and visual grounding that decodes whole bounding boxes in parallel.

ND

Not Diamond

API · Not Diamond

2

Intelligent model router API that picks the model and reasoning effort per step for coding agents to cut inference cost.

MI

mini-SWE-agent

Library

1

Minimal open-source AI software engineering agent in about 100 lines of Python that uses only bash.