🚨 China’s Ernie 4.5 Just Crushed GPT-4 🇨🇳💥
Baidu dropped Ernie 4.5 — a 424B parameter open-source model that’s beating GPT-4 in key areas, and it’s 100% free.
- Outperforms GPT-4.0 in:
- DocVQA: 93.1 vs 81.0
- ChartQA: 86.8 vs ~80
- Math (GSM8K, CMATH, MATH500):96%+
- Reasoning (BBH): 94.3 vs GPT’s 91.0
⚡ Uses 47B active parameters → faster + cheaper than GPT
🧠 Has a thinking mode like GPT-4 Turbo
🎥 True multimodal: handles text, images, audio, and video
✅ Fully open-source → run locally, fine-tune, no API fees
📥 Available now on HuggingFace
Hugging Face | GitHub | Baidu Blog | Ernie Bot (ChatGPT‑style interface)
#baidu #opensource #huggingface #china #transformers #nlp
Baidu dropped Ernie 4.5 — a 424B parameter open-source model that’s beating GPT-4 in key areas, and it’s 100% free.
- Outperforms GPT-4.0 in:
- DocVQA: 93.1 vs 81.0
- ChartQA: 86.8 vs ~80
- Math (GSM8K, CMATH, MATH500):96%+
- Reasoning (BBH): 94.3 vs GPT’s 91.0
⚡ Uses 47B active parameters → faster + cheaper than GPT
🧠 Has a thinking mode like GPT-4 Turbo
🎥 True multimodal: handles text, images, audio, and video
✅ Fully open-source → run locally, fine-tune, no API fees
📥 Available now on HuggingFace
Hugging Face | GitHub | Baidu Blog | Ernie Bot (ChatGPT‑style interface)
#baidu #opensource #huggingface #china #transformers #nlp
👾45❤4
The Core: Apple is considering using Anthropic’s Claude or OpenAI’s ChatGPT to power the next-gen Siri, moving away from its own LLMs—after Siri’s upgrade was delayed to 2026 .
- They've asked both companies to create custom models to run on Apple Cloud for internal testing .
- Project led by Mike Rockwell (Vision Pro), still evaluating Apple-built “LLM Siri” .
- Anthropic reportedly wants a multi‑billion‑dollar licensing deal—Apple using OpenAI as leverage .
Why it matters:
Siri is lagging behind Google Assistant & Alexa in AI capabilities .
Bringing in third-party AI marks a major departure for Apple’s traditionally in-house approach .
🏷️ Suggested hashtags:
#Openai #anthropic #claude #apple
- They've asked both companies to create custom models to run on Apple Cloud for internal testing .
- Project led by Mike Rockwell (Vision Pro), still evaluating Apple-built “LLM Siri” .
- Anthropic reportedly wants a multi‑billion‑dollar licensing deal—Apple using OpenAI as leverage .
Why it matters:
Siri is lagging behind Google Assistant & Alexa in AI capabilities .
Bringing in third-party AI marks a major departure for Apple’s traditionally in-house approach .
🏷️ Suggested hashtags:
#Openai #anthropic #claude #apple
📰 Everything else in AI today
RAISE Summit in Paris, July 8-9 — All things AI. Join SambaNova at booth #9, snag an invite to an exclusive soirée, and catch CEO Rodrigo Liang’s keynote on Open Source AI.*
Mark Zuckerberg introduced “Meta Superintelligence Labs” to employees, with Alexandr Wang and Nat Friedman leading 11 hires from OpenAI, Google, and Anthropic.
Apple is reportedly considering leveraging AI from Anthropic and OpenAI for the revamped Siri over in-house models, according to a new report from Bloomberg.
The Mayo Clinic unveiled StateViewer, an AI tool that analyzes brain scans to help identify nine different types of dementia at 2x the speed and 3x the accuracy.
Cursor launched new apps for mobile and browser, allowing users to manage and monitor agents via natural language outside of its IDE.
Google announced Gemini in Classroom, a suite of AI features and tools for educators for tasks like lesson planning, NotebookLM access, and student performance analytics.
Canada dropped its planned digital tax on U.S. tech giants after Trump threatened to halt trade talks.
Tinder now requires all new users in California to verify their profiles with facial recognition technology.
Grammarly acquired email efficiency tool Superhuman, Reuters reports, as part of its strategy to develop an AI-powered productivity suite and expand its business.
Apple finally scored its first major box office hit with F1, the Brad Pitt-led racing drama that debuted at No. 1 and has so far raked in $144M worldwide.
Mira Murati’s AI startup, Thinking Machines Lab, is reportedly offering exceptionally high base salaries, with some roles reaching $500K, to draw in top talent.
OpenAI said it has no current plans to deploy Google’s in-house Tensor Processing Units at commercial scale, clarifying that the chips are only being used in early testing.
President Donald Trump said that a group of “very wealthy people” is set to buy video app TikTok, banned in the U.S. due to security risks.
Biopharma company Vertex Pharmaceuticals found that one infusion of a stem cell-based treatment may have cured 10 out of 12 people with severe type 1 diabetes.
Anker announced a second product recall this month for five more power bank models because of potential risks of fire.
Joby Aviation delivered its first production eVTOL aircraft to Dubai, where it plans to launch a commercial air taxi service in early 2026.
#Openai #anthropic #Google #opensource #tiktok #claude #microsoft #huggingface
#trump
RAISE Summit in Paris, July 8-9 — All things AI. Join SambaNova at booth #9, snag an invite to an exclusive soirée, and catch CEO Rodrigo Liang’s keynote on Open Source AI.*
Mark Zuckerberg introduced “Meta Superintelligence Labs” to employees, with Alexandr Wang and Nat Friedman leading 11 hires from OpenAI, Google, and Anthropic.
Apple is reportedly considering leveraging AI from Anthropic and OpenAI for the revamped Siri over in-house models, according to a new report from Bloomberg.
The Mayo Clinic unveiled StateViewer, an AI tool that analyzes brain scans to help identify nine different types of dementia at 2x the speed and 3x the accuracy.
Cursor launched new apps for mobile and browser, allowing users to manage and monitor agents via natural language outside of its IDE.
Google announced Gemini in Classroom, a suite of AI features and tools for educators for tasks like lesson planning, NotebookLM access, and student performance analytics.
Canada dropped its planned digital tax on U.S. tech giants after Trump threatened to halt trade talks.
Tinder now requires all new users in California to verify their profiles with facial recognition technology.
Grammarly acquired email efficiency tool Superhuman, Reuters reports, as part of its strategy to develop an AI-powered productivity suite and expand its business.
Apple finally scored its first major box office hit with F1, the Brad Pitt-led racing drama that debuted at No. 1 and has so far raked in $144M worldwide.
Mira Murati’s AI startup, Thinking Machines Lab, is reportedly offering exceptionally high base salaries, with some roles reaching $500K, to draw in top talent.
OpenAI said it has no current plans to deploy Google’s in-house Tensor Processing Units at commercial scale, clarifying that the chips are only being used in early testing.
President Donald Trump said that a group of “very wealthy people” is set to buy video app TikTok, banned in the U.S. due to security risks.
Biopharma company Vertex Pharmaceuticals found that one infusion of a stem cell-based treatment may have cured 10 out of 12 people with severe type 1 diabetes.
Anker announced a second product recall this month for five more power bank models because of potential risks of fire.
Joby Aviation delivered its first production eVTOL aircraft to Dubai, where it plans to launch a commercial air taxi service in early 2026.
#Openai #anthropic #Google #opensource #tiktok #claude #microsoft #huggingface
#trump
This media is not supported in your browser
VIEW IN TELEGRAM
🚀 @Skywork_ai — All Your Work, Done by AI
Skywork helps you write, research, and create content like slides, videos, and podcasts — all from a single prompt.
🧠 What Can Skywork_ai Do?
• 📝 Generate
– Polished documents
– Structured slide decks
– Clean, analyzed spreadsheets
• 🌐 Create:
— Webpages from plain prompts
— Podcasts with real voiceovers
— Videos via SkyReels‑V2 (open-source, unlimited duration)
• 🔍 Research:
— Extract insights from PDFs, audio, video, images
— Source-cited summaries & answers
— Benchmark-topping accuracy (GAIA Score: 82.4)
• 💡 Other Features:
— Clarification Card: asks for context before working
— Multimodal Agent System: specialized + general AI agents
— Open-source AI models on Hugging Face
— Citation tracking to avoid hallucinations.
Open Skywork | @CoreAti
#SkyWork #agenticai
Skywork helps you write, research, and create content like slides, videos, and podcasts — all from a single prompt.
🧠 What Can Skywork_ai Do?
• 📝 Generate
– Polished documents
– Structured slide decks
– Clean, analyzed spreadsheets
• 🌐 Create:
— Webpages from plain prompts
— Podcasts with real voiceovers
— Videos via SkyReels‑V2 (open-source, unlimited duration)
• 🔍 Research:
— Extract insights from PDFs, audio, video, images
— Source-cited summaries & answers
— Benchmark-topping accuracy (GAIA Score: 82.4)
• 💡 Other Features:
— Clarification Card: asks for context before working
— Multimodal Agent System: specialized + general AI agents
— Open-source AI models on Hugging Face
— Citation tracking to avoid hallucinations.
Open Skywork | @CoreAti
#SkyWork #agenticai
👾46❤3
🎮 Runway Enters AI Gaming
Runway, known for AI in film, is now letting users create games with generative AI.
• New tool supports text & image prompts; full game generation coming soon
• Gaming studios are in talks to use Runway’s tech
• AI could speed up game development by 40%
#Runway #AI #Gaming #GameDev
Runway, known for AI in film, is now letting users create games with generative AI.
• New tool supports text & image prompts; full game generation coming soon
• Gaming studios are in talks to use Runway’s tech
• AI could speed up game development by 40%
#Runway #AI #Gaming #GameDev
🎬 Automate YouTube content creation with AI
The CORE: In this tutorial, you will learn how to use Google's Gemini AI to analyze videos and automatically generate noscripts, denoscriptions, tags, and chapters — streamlining your YouTube content workflow.
Step-by-step:
1. Go to Google Gemini, click thE attachment button, and upload your video file
2. Generate noscripts with: “Analyze this video and provide 10 compelling YouTube noscripts”
3. Ask it to provide you with a detailed video denoscription with a hook, summary, and call-to-action.
4. Request it for tags as comma-separated values, plus chapter timestamps
Pro tip: Save these prompts as templates and customize the tone (professional, casual, energetic) to match your brand voice.
#Google #gemini #nlp
The CORE: In this tutorial, you will learn how to use Google's Gemini AI to analyze videos and automatically generate noscripts, denoscriptions, tags, and chapters — streamlining your YouTube content workflow.
Step-by-step:
1. Go to Google Gemini, click thE attachment button, and upload your video file
2. Generate noscripts with: “Analyze this video and provide 10 compelling YouTube noscripts”
3. Ask it to provide you with a detailed video denoscription with a hook, summary, and call-to-action.
4. Request it for tags as comma-separated values, plus chapter timestamps
Pro tip: Save these prompts as templates and customize the tone (professional, casual, energetic) to match your brand voice.
#Google #gemini #nlp
👾45❤4
🚨Breaking: Higgsfield SOUL is now FREE. (Open-source)
It democratizes creativity, letting anyone make Pinterest-worthy visuals instantly.
https://higgsfield.ai/soul
It democratizes creativity, letting anyone make Pinterest-worthy visuals instantly.
https://higgsfield.ai/soul
#texttoimage #Pinterest #Higgsfield
👾45❤6
Content creation will NEVER be the same again
I just found an AI tool that’s like ChatGPT + Veo3 + Higgsfield combined
Here are 7 WILD examples 👇
Incomments .
#ChatGPT #Veo3 #Higgsfield #MidJourney
I just found an AI tool that’s like ChatGPT + Veo3 + Higgsfield combined
Here are 7 WILD examples 👇
In
#ChatGPT #Veo3 #Higgsfield #MidJourney
👾45❤4
Google Maps just became an AI BEAST and 99% of users have no clue.
Here are 12 insane AI features you MUST see 👇
#google #googlemaps #useful
Here are 12 insane AI features you MUST see 👇
#google #googlemaps #useful
👾46❤4
🗣 AI voices with personality
The Core: Rime offers the most realistic and expressive AI voices on the market today, creating agentic experiences that sound like everyday people, not robots or voice actors — while driving double digit increases in conversion for brands like Dominos.
Use Rime’s platform to:
- Deliver natural voice AI agents that laugh, breathe, sigh, and sound human
- Easily integrate multilingual text-to-speech (TTS) via API or on-prem
- Start building today with a generous free tier
Try Rime’s live chat (or +1-662-727-8948)
The Core: Rime offers the most realistic and expressive AI voices on the market today, creating agentic experiences that sound like everyday people, not robots or voice actors — while driving double digit increases in conversion for brands like Dominos.
Use Rime’s platform to:
- Deliver natural voice AI agents that laugh, breathe, sigh, and sound human
- Easily integrate multilingual text-to-speech (TTS) via API or on-prem
- Start building today with a generous free tier
Try Rime’s live chat (or +1-662-727-8948)
👾45
🫂 Sakana AI teaches models to team up
The Core: Japanese lab Sakana AI just introduced AB-MCTS, a new algorithm that lets competing AI models collaborate on complex problems, using their collective intelligence to solve problems too difficult for any of the individual models.
The details:
- The system combines ChatGPT, Gemini, and DeepSeek using adaptive search, solving 30% of ARC-AGI-2 puzzles versus just 23% for top solo models.
- AB-MCTS dynamically allocates different models based on strengths, with some handling strategy while others excel at code within the same problem.
- Researchers discovered models could build on each other's mistakes, with one model correcting flawed answers from another to reach correct solutions.
Bonus: The framework, TreeQuest, is open-source for devs to build their own collaborative AI setups.
The Core: Japanese lab Sakana AI just introduced AB-MCTS, a new algorithm that lets competing AI models collaborate on complex problems, using their collective intelligence to solve problems too difficult for any of the individual models.
The details:
- The system combines ChatGPT, Gemini, and DeepSeek using adaptive search, solving 30% of ARC-AGI-2 puzzles versus just 23% for top solo models.
- AB-MCTS dynamically allocates different models based on strengths, with some handling strategy while others excel at code within the same problem.
- Researchers discovered models could build on each other's mistakes, with one model correcting flawed answers from another to reach correct solutions.
Bonus: The framework, TreeQuest, is open-source for devs to build their own collaborative AI setups.
❤1
🧠 Scientists build an AI that can think like humans
The Core: Helmholtz Munich researchers built Centaur, an AI model simulating human decision-making with high accuracy using millions of psychology experiment choices.
Details:
- Centaur fine-tuned Meta’s LLaMA on data from 60k people, 160 experiments.
- It predicts human behavior in tasks like gambling, memory, problem-solving — even unseen ones.
- Outperformed 14 cognitive models in 31/32 tasks.
- Acts as a "virtual lab" for testing cognitive and mental health theories.
Why it matters:
- Human decisions may be more predictable than believed.
- Boosts research speed and could enable ASI-level scenario simulations.
The Core: Helmholtz Munich researchers built Centaur, an AI model simulating human decision-making with high accuracy using millions of psychology experiment choices.
Details:
- Centaur fine-tuned Meta’s LLaMA on data from 60k people, 160 experiments.
- It predicts human behavior in tasks like gambling, memory, problem-solving — even unseen ones.
- Outperformed 14 cognitive models in 31/32 tasks.
- Acts as a "virtual lab" for testing cognitive and mental health theories.
Why it matters:
- Human decisions may be more predictable than believed.
- Boosts research speed and could enable ASI-level scenario simulations.
👾46❤2🔥2🤯1
📰 Everything else in AI today
Perplexity launched Max, a new $200/mo tier giving users unlimited access to its Labs tools, early access to new products like its Comet browser and advanced models.
OpenAI is expanding its Stargate partnership with Oracle, renting about 4.5 GW of data center capacity to power its AI energy needs.
Anthropic is reportedly on pace for $4B in annual revenue, 4x higher than its projections at the start of 2025.
Google DeepMind CEO Demis Hassabis hinted at potential “playable world” models coming for its Veo 3 video generation model in a response on X.
Chinese tech giant Huawei open-sourced several of its Pangu models and the underlying reasoning tech, trained using the company’s own Ascend chips.
AI startup Lovable is reportedly set to raise a new $150M funding round, valuing the vibe-coding platform at close to $2B.
Perplexity launched Max, a new $200/mo tier giving users unlimited access to its Labs tools, early access to new products like its Comet browser and advanced models.
OpenAI is expanding its Stargate partnership with Oracle, renting about 4.5 GW of data center capacity to power its AI energy needs.
Anthropic is reportedly on pace for $4B in annual revenue, 4x higher than its projections at the start of 2025.
Google DeepMind CEO Demis Hassabis hinted at potential “playable world” models coming for its Veo 3 video generation model in a response on X.
Chinese tech giant Huawei open-sourced several of its Pangu models and the underlying reasoning tech, trained using the company’s own Ascend chips.
AI startup Lovable is reportedly set to raise a new $150M funding round, valuing the vibe-coding platform at close to $2B.
👾45❤2❤🔥1🔥1