What a week - we've got top models from 4 of the biggest AI labs in a single week: Gemini 3.8 Flash, Claude Fable 5.1, Muse Spark 1.3, GPT-6 Astra!
> A full recap is coming in a couple of hours in TestingCatalog's newsletter.
>β΄οΈ Bonus! A secret Muse agent waitlist join link is included.
Opt in below if you haven't yet π©
https://www.testingcatalog.com/#/portal/signup
> A full recap is coming in a couple of hours in TestingCatalog's newsletter.
>
Opt in below if you haven't yet π©
https://www.testingcatalog.com/#/portal/signup
Please open Telegram to view this post
VIEW IN TELEGRAM
TestingCatalog AI News
TestingCatalog | AI News & Rumours
TestingCatalog covers latest AI news, model releases, leaks, and rumours from the world of artificial intelligence
β€4π₯3 3π1
Visiting IFA 2026 Berlin today!
IRL testing time π
IFA is a major annual Berlin trade show where global brands unveil new consumer electronics, home appliances, and future technology.
IRL testing time π
β€9π2π1
IFM releases K2 Horizon: 6 models with full training record
IFM released K2 Horizon, six Apache 2.0 language models from 0.9B to 375B parameters, with training code, logs, checkpoints, evaluations, and documented reward-hacking cases, emphasizing reproducibility and auditability.
π #sponsored @testingcatalog
IFM released K2 Horizon, six Apache 2.0 language models from 0.9B to 375B parameters, with training code, logs, checkpoints, evaluations, and documented reward-hacking cases, emphasizing reproducibility and auditability.
π #sponsored @testingcatalog
TestingCatalog AI News
IFM releases K2 Horizon: 6 models with full training record
The Institute of Foundation Models published six K2 Horizon models, ranging from 0.9B to 375B, along with code, checkpoints, logs, and evaluations.
π1
Meta's Muse agent enters closed alpha with invite codes
Metaβs Muse AI agent is nearing launch in closed alpha, with waitlist and invite codes, mobile and web access, app connectors, spending controls, and task automation across email, calendar, bookings, budgets, and planning.
π #meta @testingcatalog
Metaβs Muse AI agent is nearing launch in closed alpha, with waitlist and invite codes, mobile and web access, app connectors, spending controls, and task automation across email, calendar, bookings, budgets, and planning.
π #meta @testingcatalog
TestingCatalog AI News
Meta's Muse agent enters closed beta with invite codes
Meta is adding user-distributed invite codes for Muse as the agent moves from internal testing to a limited closed alpha on mobile and web.
β€2π1
π¨ AI News | TestingCatalog
Meta's Muse agent enters closed alpha with invite codes Metaβs Muse AI agent is nearing launch in closed alpha, with waitlist and invite codes, mobile and web access, app connectors, spending controls, and task automation across email, calendar, bookingsβ¦
Early look at the UI of the upcoming Muse agent from Meta.
> According to the App Store listing, the Muse agent might act as a proactive assistant that helps users track their goals and stay on top of their schedule and tasks.
> The Muse agent transitioned from internal to closed / limited alpha testing.
> New functionality for invite codes has been added.
Soon π
> According to the App Store listing, the Muse agent might act as a proactive assistant that helps users track their goals and stay on top of their schedule and tasks.
> The Muse agent transitioned from internal to closed / limited alpha testing.
> New functionality for invite codes has been added.
Soon π
Gemini Immersive View appears in the "Students" section
Googleβs Gemini student app adds Immersive View, an image-based topic explorer with clickable nodes and layered detail. Access appears live via direct URL, though Google has not clarified if this is a test or a wider rollout.
h/t @Thomasguka
π #google @testingcatalog
Googleβs Gemini student app adds Immersive View, an image-based topic explorer with clickable nodes and layered detail. Access appears live via direct URL, though Google has not clarified if this is a test or a wider rollout.
h/t @Thomasguka
π #google @testingcatalog
TestingCatalog AI News
Gemini Immersive View appears in the "Students" section
A new Immersive View is appearing in Googleβs Gemini student app, offering image-based topic maps with clickable nodes, though rollout status is unclear.
π¨ AI News | TestingCatalog
Gemini Immersive View appears in the "Students" section Googleβs Gemini student app adds Immersive View, an image-based topic explorer with clickable nodes and layered detail. Access appears live via direct URL, though Google has not clarified if this isβ¦
Media is too big
VIEW IN TELEGRAM
GOOGLE π₯: Gemini now offers a new Immersive View experience in the Students tab!
> Users can pick from a variety of topics to explore, starting with dinosaurs and continuing with space and rockets.
> Each experience opens an interactive image where users can drill down into pre-selected details. In some cases, it can go quite deep, even explaining molecular structure.
I guess my 5yo will have some testing time today! It is pretty cool to explore.
h/t @Thomasguka
> Users can pick from a variety of topics to explore, starting with dinosaurs and continuing with space and rockets.
> Each experience opens an interactive image where users can drill down into pre-selected details. In some cases, it can go quite deep, even explaining molecular structure.
I guess my 5yo will have some testing time today! It is pretty cool to explore.
h/t @Thomasguka
β€5 5π3
OpenAI prepares managed agents for DevDay 2026
OpenAIβs DevDay may center on managed agents for developers and businesses, echoing Anthropicβs model. Expected reveals include model and security updates, plus agent-based ads that could challenge current digital marketing.
π #openai @testingcatalog
OpenAIβs DevDay may center on managed agents for developers and businesses, echoing Anthropicβs model. Expected reveals include model and security updates, plus agent-based ads that could challenge current digital marketing.
π #openai @testingcatalog
TestingCatalog AI News
OpenAI prepares managed agents for DevDay 2026
OpenAI appears to be preparing managed agents for DevDay, including configurable environments, skills, plugins, and a business-focused experience.
β€2π1
π¨ AI News | TestingCatalog
OpenAI prepares managed agents for DevDay 2026 OpenAIβs DevDay may center on managed agents for developers and businesses, echoing Anthropicβs model. Expected reveals include model and security updates, plus agent-based ads that could challenge current digitalβ¦
Media is too big
VIEW IN TELEGRAM
OPENAI π₯: As tradition dictates, OpenAI plans to announce a new solution for its agent ecosystem at the upcoming DevDay 2026.
This time, we will get Managed Agents π€
> Users will be able to create Agents, Environments, and Agent Sessions directly via the OpenAI Platform.
> The Codex Plugin will help users build, configure, and deploy agents, lowering the barrier to entry.
> OpenAI's Managed Agents could be hosted on the OpenAI platform or self-hosted.
Loads of details inside π
This time, we will get Managed Agents π€
> Users will be able to create Agents, Environments, and Agent Sessions directly via the OpenAI Platform.
> The Codex Plugin will help users build, configure, and deploy agents, lowering the barrier to entry.
> OpenAI's Managed Agents could be hosted on the OpenAI platform or self-hosted.
Loads of details inside π
Anthropic is adding a dedicated button for viewing usage details to the chat as well as a new customisation section for Claude Code on Claude for iOS.
β€10π₯3 1
Google tests proper Projects for the Gemini desktop app
Google is internally testing a Projects feature for Gemini desktop: persistent workspaces with custom instructions, files, and grouped chats. It mirrors long-standing ChatGPT and Claude setups, but no public launch details are confirmed.
π #google @testingcatalog
Google is internally testing a Projects feature for Gemini desktop: persistent workspaces with custom instructions, files, and grouped chats. It mirrors long-standing ChatGPT and Claude setups, but no public launch details are confirmed.
π #google @testingcatalog
TestingCatalog AI News
Google tests proper Projects for the Gemini desktop app
Google is internally testing Projects for the Gemini desktop app, grouping custom instructions, files and related chats in dedicated workspaces.
π1
π¨ AI News | TestingCatalog
Google tests proper Projects for the Gemini desktop app Google is internally testing a Projects feature for Gemini desktop: persistent workspaces with custom instructions, files, and grouped chats. It mirrors long-standing ChatGPT and Claude setups, but noβ¦
This media is not supported in your browser
VIEW IN TELEGRAM
Google is healing? β€οΈβπ©Ή
A canonical Projects implementation has been spotted in testing on the Gemini Desktop app.
> The current implementation looks very similar to Projects on Gemini Business.
> This experience is being tested alongside NotebookLM integration and unlikely meant to replace it.
First, we've got custom MCPs on Gemini Business, and now this?
A canonical Projects implementation has been spotted in testing on the Gemini Desktop app.
> The current implementation looks very similar to Projects on Gemini Business.
> This experience is being tested alongside NotebookLM integration and unlikely meant to replace it.
First, we've got custom MCPs on Gemini Business, and now this?
β€7π3 3
Mistral AI raised β¬3 billion in a Series D funding at more than β¬21 billion in a post-money valuation.
The biggest EU Round ever! π€―
"Samsung Electronics led the round, joined by co-leads Scaleup Europe Fund, managed by EQT, and existing investor PSG Equity."
The biggest EU Round ever! π€―
β€12π1
π¨ AI News | TestingCatalog
DAILY AI BRIEF π β Sept 5 OPENAI π₯ - Astra is rolled out to all Plus and Business β not just Pro / Enterprise. API live at $10 / $50. GPT-6 Pro in Chat is Astra. - Full banked reset today for Plus, Pro, and Business. Lands EOD. Create or upgrade before 8pmβ¦
DAILY AI BRIEF π β Sept 8
META π₯:
β’ First look at Muse (Hatch) mobile UI: proactive agent for goals, calendar, email, bookings, and spend, now in closed beta with invite codes.
GOOGLE π₯:
β’ Immersive View in Gemini Students β image topic maps with clickable nodes. Worth testing!
OPENAI π₯:
β’ Writing Style for ChatGPT Work learns your voice from Gmail, Drive, Slack, and SharePoint.
β’ Usage-reset week: banked Codex resets plus a Sept 7 Plus/Pro/Business flush that is not bankable.
β’ DevDay prep: managed agents on the platform (environments, skills, plugins). Sept 29.
MISTRAL π₯:
β’ β¬3B Series D at >β¬21B, Europeβs largest tech equity round, led by Samsung with Scaleup Europe Fund and PSG.
ANTHROPIC π₯:
β’ Walked away from the ~$6B Decart deal after diligence; collaboration still possible.
Anything to add here? π
META π₯:
β’ First look at Muse (Hatch) mobile UI: proactive agent for goals, calendar, email, bookings, and spend, now in closed beta with invite codes.
GOOGLE π₯:
β’ Immersive View in Gemini Students β image topic maps with clickable nodes. Worth testing!
OPENAI π₯:
β’ Writing Style for ChatGPT Work learns your voice from Gmail, Drive, Slack, and SharePoint.
β’ Usage-reset week: banked Codex resets plus a Sept 7 Plus/Pro/Business flush that is not bankable.
β’ DevDay prep: managed agents on the platform (environments, skills, plugins). Sept 29.
MISTRAL π₯:
β’ β¬3B Series D at >β¬21B, Europeβs largest tech equity round, led by Samsung with Scaleup Europe Fund and PSG.
ANTHROPIC π₯:
β’ Walked away from the ~$6B Decart deal after diligence; collaboration still possible.
Anything to add here? π
π¨ AI News | TestingCatalog
OPENAI π₯: As tradition dictates, OpenAI plans to announce a new solution for its agent ecosystem at the upcoming DevDay 2026. This time, we will get Managed Agents π€ > Users will be able to create Agents, Environments, and Agent Sessions directly via theβ¦
OpenAI is planning to make managed agents easily accessible in a way that anyone who uses Codex can just ask AI to build it, configure and deploy.
Sample use cases π
> Research Agent - Searches the web and connected data sources to produce structured research.
> Customer Support Agent - Resolves support requests, retrieves account information, and escalates.
> DevOps Assistant - Monitors deployments, investigates incidents, and helps troubleshoot.
> Meeting Assistant - Transcribes meetings, captures action items, and generates summaries.
> Analytics Agent - Queries data sources, generates reports, and explains trends in natural language.
> Analytics Agent - Queries data sources, generates reports, and explains trends in natural language.
Sample use cases π
> Research Agent - Searches the web and connected data sources to produce structured research.
> Customer Support Agent - Resolves support requests, retrieves account information, and escalates.
> DevOps Assistant - Monitors deployments, investigates incidents, and helps troubleshoot.
> Meeting Assistant - Transcribes meetings, captures action items, and generates summaries.
> Analytics Agent - Queries data sources, generates reports, and explains trends in natural language.
> Analytics Agent - Queries data sources, generates reports, and explains trends in natural language.
Duliffusion models are evolving π
Inception launched Mercury 2.5 diffusion model and it performs on par with Gemini 3.5 Flash and GPT 5.6 Luna Low at a much faster speed.
Inception launched Mercury 2.5 diffusion model and it performs on par with Gemini 3.5 Flash and GPT 5.6 Luna Low at a much faster speed.
Intelligence: +40% vs Mercury 2
Speed: 1,107 tok/s on widely available NVIDIA GPUs
β€8π4π1π΄1