- AI Fire
- Posts
- 🤖 Chinese Mode Takes On o1, Again?
In partnership with
Plus: Steer Your Chromebook with a Nod
Read time: 5 minutes
ByteDance is hot on OpenAI’s heels with the launch of Doubao-1.5-Pro, claiming superior performance to o1 on AIME benchmarks. Earlier this week, DeepSeek’s R1 also outshined o1 in multiple tests.
What are on FIRE 🔥
IN PARTNERSHIP WITH HUBSPOT
Stay Competitive: Embrace AI or Get Left Behind
In today’s fast-paced world, adopting AI is no longer optional for small businesses looking to stay competitive. HubSpot’s AI-powered customer platform is your key to staying ahead of the curve. With HubSpot, you can automate those manual tasks that slow you down, personalize every customer interaction, and unlock powerful, actionable insights from your data. You’ll make smarter, data-driven decisions to scale your business effectively and optimize your strategies. Why Choose HubSpot for Your Business?
Affordable at $20/Month: Access marketing, sales, and customer support all in one – cheaper than buying separate products.
Centralized Data: Eliminate silos by keeping all your data in one unified location.
Automation: Free yourself from tedious tasks and focus on the more impactful areas of your business.
Smart CRM: Gain insights about your customers that your competitors can only dream of. Don't fall behind: Leverage AI to propel your business forward.
AI INSIGHTS
🤖 China’s AI Push: ByteDance Enters the Chat
ByteDance made waves in the AI space with the release of Doubao 1.5 Pro. The numbers? Jaw-dropping. The performance? Right up there with the best.
From ByteDance’s AI Drop
Doubao 1.5 Pro is here, and it’s cheap. Really cheap.
$0.022/M cached input tokens, $0.11/M input, and $0.275/M output.
That’s 50x cheaper than GPT-4o and 5x cheaper than DeepSeek.
The model’s sparse MoE architecture enables 32k and 256k context, giving it the ability to handle vast amounts of data with ease.
Performance on Benchmarks
Doubao doesn’t just talk the talk—it walks the walk:
It outperformed GPT-4o, Claude 3.5 Sonnet, and DeepSeek V3 across coding, reasoning, knowledge, and Chinese language benchmarks.
And ByteDance didn’t stop there. They also open-sourced veRL, the reinforcement learning library that powers Doubao’s reasoning abilities.
Oh, and there’s more. Meet UI-TARS—ByteDance’s new open-source GUI AI agent. It uses screenshots as input to reason and perform computer interactions, bringing even more versatility to the table.
Why it matters: Even with chip restrictions and the TikTok ban drama, ByteDance is showing no signs of slowing down. The Doubao 1.5 Pro launch, alongside open-source tools like veRL and UI-TARS, highlights how China’s AI scene isn’t just keeping pace—it’s pushing boundaries.
As AI becomes "too cheap to meter," the race to lead the AI arms race is getting fiercer. Doubao 1.5 Pro might just be the model that sets a new standard for affordability and efficiency.
2025 is shaping up to be the year we find out if anyone can keep up.
PRESENTED BY MORNING BREW
The gold standard of business news
Morning Brew is transforming the way working professionals consume business news.
They skip the jargon and lengthy stories, and instead serve up the news impacting your life and career with a hint of wit and humor. This way, you’ll actually enjoy reading the news—and the information sticks.
Best part? Morning Brew’s newsletter is completely free. Sign up in just 10 seconds and if you realize that you prefer long, dense, and boring business news—you can always go back to it.
🎁 Today's Trivia - Vote, Learn & Win!
Win a 3-month membership at AI Fire Academy (500+ AI Workflows, AI Tutorials, AI Case Studies) just by answering the poll.
Which Is NOT a ChatGPT Add-in Feature? |
TODAY IN AI
AI HIGHLIGHTS
💡 OpenAI released Operator, an AI agent that performs tasks like browsing, clicking, and interacting with web pages. Available for U.S. Pro users as a research preview.
💥 Elon Musk and Sam Altman had a back-and-forth on X, with a mix of jabs and a rare compliment. Musk claimed OpenAI and its partners might not have enough funds to pull off their $500B infrastructure deal.
🖥️ Google is rolling out facial controls for Chromebooks, designed for people with motor impairments. This isn’t Google’s first time using face-as-cursor tech.
📱 Samsung’s new phones bring smarter AI with Gemini upgrades, context-aware interactions, and multimodal agent features, making the S25 series stand out.
🚀 At Davos, OpenAI CPO Kevin Weil shared that work has already started on the successor to the o3 reasoning model, promising even greater capabilities ahead.
DAILY AI FUNDRAISING
Former Google VP Maria Zhang launches Palona AI with $10M in seed funding. The startup creates smart AI agents for retail, offering personalized, action-oriented customer interactions to boost business growth.
AI SOURCES FROM AI FIRE
NEW EMPOWERED AI TOOLS
💻 Lovable + Builder.io turns Figma designs into full-stack apps with AI
📝 MeetMinutes transforms your pre- and post-meeting workflows
📄 UPDF AI Chat with any PDF, ask AI anything, summarize or translate
🛠️ Supabase AI Assistant helps from idea to Postgres database in no time
📊 Equals is the fastest and most flexible way to report on your SaaS business
AI QUICK HITS
AI CHART
Last year, we gathered insights from over 1,200 companies to understand where AI is heading. Spoiler: It’s not as futuristic as it seems.
Here’s the real story of AI in 2025: most companies are still figuring things out.
From the AI Adoption Report
52.2% of businesses are relying on internal tools for AI development.
But major gaps—like evaluation and monitoring—are holding progress back.
Hallucinations are the top challenge, with 57.4% of companies struggling to keep their models on track.
And the tooling? Still catching up. Only 38% of companies use automated evaluation tools, meaning manual testing is still king.
What Companies Are Building
Document parsing is leading the way (59.7% adoption), followed by customer service chatbots (51.4%) and natural language analytics (43.8%).
Other cool trends: Retrieval-Augmented Generation (RAG) and vector databases are becoming must-haves for accuracy, used by 59.7% of teams.
But here’s the kicker: 24.2% of companies report “no measurable impact yet.” Looks like early-stage efforts still dominate the space.
What’s Holding Everyone Back?
Hallucinations are the #1 issue (57.4%).
Teams are also struggling to prioritize impactful use cases (42.5%) and deal with skill gaps.
Evaluation and monitoring remain a bottleneck—most companies (55.3%) rely on in-house tools, but scalability is still a question mark.
The intelligence is there, but making it practical? That’s the hard part.
Looking Ahead: 2025 Goals
Companies are dreaming big:
58.8% want to build more customer-facing use cases.
55.2% are exploring agentic workflows (AI making complex decisions).
Upskilling is on the radar too, with 41.9% planning to train their teams.
If 2024 was about jumping on the AI bandwagon, 2025 is about learning to steer the wheel.
And Speaking of 2025
The future of AI hinges on better tools. Companies investing in evaluation, monitoring, and smarter workflows will set the pace. The big shift to agentic workflows shows a hunger for AI that’s not just smart—but game-changing.
AI isn’t taking over just yet, but the race to make it practical is ongoing.
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