AI in week 38.

JE

Jev

Model · Type Safe AI

4

Non-chat "decision model" from Type Safe AI that returns calibrated probabilities over predefined answer options.

MV

Mimo V2.6

Model · Xiaomi

2

Xiaomi's MiMo-V2.6 model series, with a 1T-parameter Pro and a 311B Flash variant, plus a Mimo Desktop harness.

Q3

Qwen 3.8 Omni Flash

Model · Alibaba

2

Alibaba's fast multimodal Qwen model with text, image, audio and video input, 1M-token context and real-time streaming.

LA

Laya

Model

1

Open-weight non-autoregressive System 1 decision model returning calibrated probabilities in one forward pass across 100+ languages.

B2

Bonsai 2 27B

Model · PrismML

1

Ternary-weight (-1, 0, +1) compression of a 27B model by PrismML, shrunk to about 5.9 GB for local use.

N3

Needle 3

Model · Cactus Compute

1

Tiny 29M–121M parameter on-device model for tool calling and extraction, shipped as an 8–29 MB binary.

BN

Bespoke Nimble 9B

Model · Bespoke Labs

1

Apache-2.0 LoRA adapter on Qwen3.5-9B that scores answer choices, booleans and rubric levels against a given context.

O1

Occamy 1.0

Model · Accio Lab

1

Open-weights 35B MoE (3B active) agentic model from Accio Lab, post-trained from Qwen3.6-35B-A3B.

ZG

ZGCM-1

Model

1

Open 7.39B dense model for math reasoning and agentic search, with a 256K context and the full training pipeline released.

Routing system from Unbiased that dispatches requests across several models with a checking harness; stealth name Union Alpha.

Geometry-guided video-to-video model from Viggle that re-renders existing footage along a new camera path and timing.

G3

Gemini 3.8 Live

Model · Google

2

Google's real-time voice model with an Extended Thinking variant, available in the Gemini app, AI Studio and Search Live.

R2

R2T2

Model

1

Open-source append-only streaming speech recognition model with configurable low-latency decoding, English and Mandarin.

Q3

Qwen 3.8 Live Translate

Model · Alibaba

1

Alibaba Qwen real-time speech interpreter that translates live audio or video with speaker labeling and voice cloning.

EM

ElevenLabs Music v2.5

Model · ElevenLabs

1

ElevenLabs' text-to-music model that generates tracks with or without vocals from natural-language prompts.

O3

Odyssey 3

Model · Odyssey

2

Odyssey 3 is an interactive world model from Odyssey that generates explorable, steerable simulated environments.

JA

Jing and Dao

Model · XGen Labs

1

Research preview from XGEN Labs: Dao, a persistent world-state engine, and Jing, a per-character experience model.

Google Labs agent for families that sends a shared daily brief and helps manage events and tasks.

CC

Claude Cowork + Chat merge

App · Anthropic

1

Anthropic merges Claude Cowork, chat and Design into one Claude app with docs, slides and scheduled background tasks.

MC

MiniMax Code CLI

Agent · MiniMax

1

Open-source terminal coding agent from MiniMax that works with MiniMax models or your own model API.

AF

Astra for Law

App · OpenAI

1

GPT-6 Astra configured for legal work, with a U.S. legal search index, legal instructions and privacy controls for law firms.

GP

GPT Policy

Research

1

Research framework letting a frozen VLM agent learn robot tasks in context from videos, images and feedback, no fine-tuning.

UH

UBTECH humanoid robot factory

Hardware · UBTECH

1

UBTECH's Liuzhou factory in China, built to produce over 10,000 Walker S and Cruzr humanoid robots per year.

DR

Dream RSI

Research · Google

1

Google research technical report on improving an agent's exploration policy by replaying past discovery trees offline.

Security research report on an AI-assisted attack chain against OpenAI systems using a libheif bug and an SSO flaw.

EN

Engram

Research · DeepSeek

1

DeepSeek research on a separate large conditional memory module for static facts, held in server RAM instead of GPU memory.

WM

We Must Pace the Frontier

Research · Anthropic

1

Essay by Dario Amodei proposing a three-step plan to pace frontier AI capability growth to match safety work.

Meta's paid subscription tier adding premium features to Instagram, Facebook and WhatsApp plus higher Meta AI usage.

CA

ChatGPT Ads (conversational ads)

App · OpenAI

1

OpenAI's advertising platform for ChatGPT, with sponsored ads, Sponsored Agents and AI tools for advertisers.

GE

God's Eye View

App · Bilawal Sidhu

1

Open-source browser app that layers live public data feeds on a 3D globe, with realtime AI voice control.