- AI Fire
- Posts
- 🦾 Current AI Won't Reach AGI?
In partnership with
Plus: Korean Model Dominates China’s Rival
Read time: 5 minutes
It seems like the trend for new AI models is not just about improving capabilities but also making them cheaper. Meanwhile, OpenAI, the biggest player, hasn’t focused on lowering prices like other new models.
What are on FIRE 🔥
PRESENTED BY SALEFORGE
Maximize Sales Pipeline Coverage with AI Agent Frank
Lower costs by letting Agent Frank take care of prospecting, crafting messages and booking meetings while your team can focus on closing deals!
Agent Frank can operate 24/7 and is fully customizable to best suit your company’s needs. By adding AI Agents to your team you can replace dozens of tools and scale your outreach without scaling your human team. Book a demo now to get a complete walkthrough:
AI INSIGHTS
🤖 LG Introduces New R1-Beating Reasoning Model
EXAONE Deep is a reasoning AI model developed by LG AI Research. It excels in understanding mathematical logic, scientific concepts, and solving programming problems.
Key takeaways:
EXAONE Deep 32B outperforms competitors (DeepSeek’s R1) in tough math benchmarks, despite being just 5% the size of other models.
Scored 94.5 points in CSAT 2025 Mathematics and 90.0 points in AIME 2024.
In AIME 2025, EXAONE Deep 32B performed on par with the much larger DeepSeek-R1 (671B model).
EXAONE Deep 7.8B and 2.4B also excelled in MATH-500 and AIME benchmarks, showing strong performance in smaller models.
Scored 66.1 on the GPQA Diamond test, proving its strong abilities in scientific reasoning (physics, chemistry, biology).
Scored 59.5 on LiveCodeBench, showcasing solid coding capabilities.
Both 7.8B and 2.4B models ranked first in GPQA Diamond and LiveCodeBench benchmarks, excelling in specialized areas.
Why it matters: In conclusion, EXAONE Deep 7.8B's performance across benchmarks, even outperforming OpenAI's o1 Mini, highlights a key shift in AI development. Not only is AI intelligence improving, but it’s also becoming smaller, more efficient, and more cost-effective. A 32B model that outperforms the much larger DeepSeek R1 in most benchmarks proves that AI is evolving rapidly—both in capability and in resource efficiency—setting a new standard for future AI models.
TODAY IN AI
AI HIGHLIGHTS
💡 OpenAI just launched o1-pro, a powerful AI model costing 10 times more than the standard o1 - $150 per million tokens for input and $600 for output.
📱 The Pixel 9a, priced at $499, features the Google Tensor G4 chip and Gemini Nano 1.0 XXS AI, but its 8GB RAM limits its AI capabilities compared to flagship models.
🏀 A $1 million March Madness bracket challenge pits AI against pro gambler Sean Perry. The goal is to show AI’s data-driven predictions can beat human guesses.
👨🏼💻 AI, like Claude, has advanced from solving basic cybersecurity challenges to tackling more complex ones, including CTF (Capture The Flag) challenges, but is still at undergraduate-level expertise and behind expert humans in tasks like reverse engineering and reconnaissance.
🤖 Zora AI, developed by Deloitte, follows Deloitte Trustworthy AI principles to ensure security, transparency, and reliability. It offers autonomous agents to improve business functions like finance, sales, and customer service.
💰 AI Daily Fundraising: Elon Musk's xAI and Nvidia have joined BlackRock, Microsoft, and MGX in a $30 billion AI fund to expand data centers and AI infrastructure across the U.S., starting with a facility in Texas by 2026.
AI SOURCES FROM AI FIRE
AI TUTORIAL
Build a fully functional app with on Bubble with AI-powered visual development
From vision to reality — in minutes
The world needs your ideas. Bubble AI empowers you to build fully functional apps with AI, even if you've never written a line of code. It could be a social app, a marketplace, or a productivity tool. Start building it today.
Describe your app, and AI will build it
Describe your app idea, and Bubble AI will generate a fully functional app in minutes, complete with a customized design, data structure, and core logic.
Easy customization. No code required.
Bubble's no-code editor lets you make changes quickly. Plus, we handle hosting and security, so you can focus on what matters: building a great product.
Launch quickly, get valuable feedback
On Bubble, you can iterate rapidly and incorporate customer feedback as it comes in. Spend less time on early drafts and more time refining your product based on real-world insights.
NEW EMPOWERED AI TOOLS
🖥️ Tencent’s Hunyuan 3D Models generate high-quality multiview shapes with 3D technology.
🎥 Stable Virtual Camera by Stability enables multi-view video generation with 3D control.
💎 Diamond AI Code Reviewer gives instant feedback for faster, better code.
👤 Aha is the world's first AI influencer marketing team
🎁 Kintsugi is AI-driven platform doesn’t just simplify sales tax; it transforms it.
AI QUICK HITS
🦾 76% of AI esearchers believe scaling up current AI approaches won't lead to AGI (Link)
📰 Italian newspaper says it has published world’s first AI-generated edition (Link)
🔎 Microsoft introduces Model Context Protocol (MCP) in Copilot Studio, simplifying integration with AI apps and agents (Link)
🤖 Noam Brown, lead researcher for AI reasoning at OpenAI said AI reasoning models could’ve emerged decades ago (Link)
😂 AI Beats Humans at Meme Humor, But Humans Still Win Best Joke (Link)
AI JOBS
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply