[AInews] @MicrosoftAI dropped their full-house of models at #Ignite2026 at a time
๐ซช Coding
| MAI Code 2: frontier-level coding at ~40% of the price
| MAI Code 2 Flash: fast & cheap
๐ซช Image
| MAI Image 2.7: create & edit
| MAI Image 2.7 Flash: images at scale
๐ซช Voice
| MAI Realtime 1: live voice
| MAI Transcribe 2: multilingual transcription that helps 10M doctorโpatient transcripts per month via #DragonCopilot (an AI clinical assistant for healthcare teams by #Microsoft)
๐ซช Cyber
| MAI Cyber 2 Flash: frontier cyber, low cost
๐ซช Yes/No Model
| Decision 1.1: 2x faster, now sees images
๐ซช Infrastructure
| Maia 200: Microsoft's AI chip
| MAI Licensed Weights: run MAI models in your own Azure
#AInews #MicrosoftAI
[AIwiki] ๐ฌ๐ผ๐'๐ฟ๐ฒ ๐ป๐ผ๐ ๐ฎ๐ป ๐๐ ๐ฒ๐ป๐๐ต๐๐๐ถ๐ฎ๐๐ ๐ถ๐ณ ๐๐ผ๐ ๐ฑ๐ผ๐ป'๐ ๐ฎ๐ฐ๐๐๐ฎ๐น๐น๐ ๐ธ๐ป๐ผ๐... how #DecisionAPI now
Been 2 months since the day @typesafeai swept the world with the #YesNo concept model at shocking $0 for output. And ofc there are many companies jumping in, Google, OpenAI and even @Cloudflare ๐ could not avoid this FOMO.
Here what the Yes-No market looks like as of Nov 2026.
#AInews #AIWiki
[AInews] weekly recap
#Anthropic introduced the twin model ๐๐น๐ฎ๐๐ฑ๐ฒ ๐๐ฎ๐ฏ๐น๐ฒ ๐ฑ.๐ฎ & ๐ ๐๐๐ต๐ผ๐ ๐ฑ.๐ฎ
| beat Opus 5.5 and are up to 30% faster. Fable is for everyone; Mythos is for verified labs.
#Google introduced ๐๐ฒ๐บ๐ถ๐ป๐ถ ๐ฐ ๐๐๐ฑ๐ฟ๐ผ๐ด๐ฒ๐ป
| fast coder that is 2.5x faster than 3.8 Flash.
Google launched the ๐๐ฒ๐บ๐ถ๐ป๐ถ ๐๐ฒ๐ฐ๐ถ๐๐ถ๐ผ๐ป๐ ๐๐ฃ๐ FINALLY
| answers up to 1,000 yes/no questions in under 250 ms with 0 output tokens billed.
Google introduced ๐๐ฒ๐บ๐ถ๐ป๐ถ ๐ฅ๐ผ๐ฏ๐ผ๐๐ถ๐ฐ๐ ๐ฎ.๐ฑ
| which learns robot skills in simulation and hit 96% success on Mercedes-Benz factory lines.
Google introduced ๐๐น๐ฝ๐ต๐ฎ๐๐ฒ๐ป๐ผ๐บ๐ฒ ๐ฎ
| able to finds nearly 2x more disease mutations, one cell type at a time.
#Alibaba previewed ๐ค๐๐ฒ๐ป๐ฐ-๐ ๐ฎ๐
| with deeper knowledge and stronger coding and agent skills.
#Meta introduced ๐ ๐๐๐ฒ ๐๐บ๐ฏ๐ฒ๐ฟ
| its smallest model, with a 1M-token context and text, image, audio and video input.
#AInews #AIrecap
[AIwiki] You're not an AI enthusiast if you don't actually know these benchmarks, which big AI companies are using to compare every new model
#AIwiki#AInews
[AInews] #Claude now right in your Google meeting. Big AI firms are non-stop killing small AI start-ups
๐ฌ Live whiteboard
| sketch the ideas with your team live right in the call
๐ฌ Ask in chat
| no hassle switching back and forth your screen
๐ฌ Notes. Actions. Recap.
| no more "who's writing this down and send over on Slack?"
Invite Claude to your next meeting or privacy concerns, no? And why not @GeminiApp introduce this feature first but Anthropic?
clip re-made by #AnnieAgent
#AInews
[AInews] Meet the #Gemini4 family by Google
๐๐๐ฑ๐ฟ๐ผ๐ด๐ฒ๐ป (โ Flash): fast and efficient for everyday coding - yesterday announced
๐๐ฎ๐ฟ๐ฏ๐ผ๐ป (โ Ultra): frontier power for the toughest problems - rolled out 2 days ago
๐๐ฟ๐ด๐ผ๐ป (โ Pro): the flagship for complex reasoning - rolled out a month ago
So if their very last model starting with K, they will have #HACK completely, right @GeminiApp ? ๐
#AInews
[AInews] #Anthropic back to the race with their strongest model #Fable 5.2 & #Mythos 5.2
Fable 5.2 & Mythos 5.2 are a step above their 5.1 versions
| generate outputs up to 30% faster
| cheaper a bit $8 in & $40 out (compared to $10 & $50)
Fable & Mythos share the same underlying model
| best for the most ambitious, open-ended technical work
| Mythos for firms and organization applications though
| Opus 5.5 is still for your daily coding and knowledge tasks
#AInews
[AInews] weekly recap Nov 2 to Nov 8 (a bit late i know ๐ช)
๐จ๏ธ #Google introduced Gemini Omni 1.2 Flash at ~$0.07/s, able to make clips up to 15s.
๐จ๏ธ #OpenAI introduced GPT-6 Cyber to study the Linux kernel with Mozilla, Palo Alto Networks, Calif, and Cisco.
๐จ๏ธ OpenAI launched ChatGPT Images 2.6, can make animated GIFs.
๐จ๏ธ #Tencent released Hy4 as the 1st Hy model with vision and computer use.
#AInews #AIresearch
[AInews] #OpenAI introduces ๐๐ต๐ฎ๐๐๐ฃ๐ง ๐๐บ๐ฎ๐ด๐ฒ๐ ๐ฎ.๐ฒ with 50% faster in speed ~ 6 seconds ๐ญ
| New features: animated GIFs, SVG exports, and 2 models GPT-Image-2.6 Flare & Sunburst
And the gif attached is of course made by Image 2.6
#AInews
[AIresearch][hacking] The #BinanceHack on Oct 29, 2026 - a full explanation for 10-year-old in 2 mins
Lesson learnt: Never ever use your company device to do any job interview out there
#AInews#AIresearch
[AInews] A Chinese streamer ๐ท๐๐๐ ๐๐ผ๐น๐ฑ ๐ฎ ๐๐ฟ๐ถ๐น๐น๐ถ๐ผ๐ป ๐๐ ๐๐ผ๐ธ๐ฒ๐ป๐ ๐ถ๐ป ๐ฏ๐ฌ ๐บ๐ถ๐ป๐๐๐ฒ๐
| Austin Li (aka the Lipstick King) sold ~100,000 monthly QwenWork plans by #Alibaba roughly 2 trillion tokens
| First time he's put an AI product on his livestream as well as first time, the AI tokens become a major category in the livestream industry
#AInews
[AIresearch][science] How @claudeai helped complete the missing map of the sky
| article: ๐ง๐ต๐ฒ ๐บ๐ถ๐๐๐ถ๐ป๐ด ๐บ๐ฎ๐ฝ ๐ผ๐ณ ๐๐ต๐ฒ ๐๐ธ๐
| published date: Oct 8, 2026
#AIresearch#AInews
[AInews] weekly recap
#Google introduce ๐๐ฒ๐บ๐ถ๐ป๐ถ ๐ฐ ๐๐ฎ๐ฟ๐ฏ๐ผ๐ป their next frontier model after Gemini 4 Argon
Google introduce ๐๐ฒ๐ป๐ถ๐ฒ ๐ฐ interactive world model
Google test 4 TPUs space, plan ๐ฃ๐ฟ๐ผ๐ฝ๏ฟฝ๏ฟฝ๏ฟฝ๏ฟฝ๐ฐ๐ ๐ฆ๐๐ป๐ฐ๐ฎ๐๐ฐ๐ต๐ฒ๐ฟ in 2027 with 2 satellites
Google ๐ฆ๐๐ป๐๐ต๐๐ detection is now in Gemini
-----
#Anthropic rename their Claude Max 5x to ๐ ๐ฎ๐ $๐ญ๐ฌ๐ฌ and Max 20x to ๐จ๐น๐๐ฟ๐ฎ $๐ฎ๐ฌ๐ฌ, drop 5-hour limit
Anthropic file ๐ฆ-๐ญ registration statement with SEC for ๐๐ฃ๐ข
-----
#OpenAI roll out ๐๐ฃ๐ง-๐ฒ.๐ญ ๐ง๐ฒ๐ฟ๐ฟ๐ฎ - a fast visual model
OpenAI give ๐๐ต๐ฎ๐๐๐ฃ๐ง ๐ฃ๐ฟ๐ผ & ๐๐๐๐ถ๐ป๐ฒ๐๐ $500 monthly API credit for 1st project
-----
#Tencent release ๐๐-๐ฆ๐ถ๐บ๐๐น๐ ๐ง speech translate fast under 3 seconds
#AInews
[AInews] @OpenAI limits how many paid ChatGPT accounts you can use on one device as well as under one payment card.
The firm wants their users to UPGRADE more existing accounts, instead of buying extra accounts to save money.
Free accounts won't be affected though.
#AInews