🤖 A 100% Open-source Model Better Than GPT-4?

AI Academy | Advertise | 88 AI Cheat Sheets | AI Mastery AZ Course

Plus: 10 Things Never Say to AI Bots

Read time: 5 minutes

Wait, so now there's a fully open-source model outperforming GPT-4? It was developed by a Chinese team. The hard truth about Chinese tech: while America pauses, they push forward - cheaper, faster, and stronger.

What are on FIRE 🔥

• 🤖 DeepSeek-V3, Ultra-large Open-source AI

• 🌟 AI Highlights

• 💸 Daily AI Fundraising

• 📚 AI Sources From AI Fire

• 🏅 AI Tools

• ⚡ 5 AI Quick Hits

• 📊 AI Chart

• 💼 4 AI Jobs

IN PARTNERSHIP WITH 100DAYSOFNOCODE & 100DAYSOFAI

Learn AI and NoCode skills for free

Learn NoCode and AI skills such as creating AI Agents or AI Marketing in just 30 min a day. It's free, fun, and effective bite-sized lessons delivered to your inbox daily. All challenges start Jan 1st, 2025

Pick Your Free Challenge ⚡️

AI MODELS

🤖 DeepSeek-V3, Ultra-large Open-source AI

A Chinese lab created DeepSeek V3, one of the most powerful open AI models, allowing developers to freely modify and use it.

Key takeaways:

• DeepSeek V3 outperforms both other "open" AI models (like Meta’s Llama 3.1 and OpenAI’s GPT-4) and "closed" models, which are accessible only through APIs.

• Trained on 14.8 trillion tokens. (1 million tokens = approximately 750,000 words).

• This is about 1.6 times the size of Meta’s Llama 3.1 405B, which has 405 billion parameters. 671 billion parameters (or 685 billion on the Hugging Face platform).

• Trained using Nvidia H800 GPUs in a data center over the course of 2 months.

• The training cost was approximately $5.5 million, which is much lower than models like GPT-4.

• Beats Llama 3.1 405b on almost every benchmark. 60 tokens/second (3x faster than V2!)

Why it matters: DeepSeek V3 is setting a new standard for open-source AI. Ranked 2nd on the Aider Polyglot leaderboard, it has proven its prowess, outperforming GPT-4o and even surpassing Claude Sonnet 3.5 in coding benchmarks.

This cutting-edge model combines power with accessibility. Developers can access the code on GitHub under an MIT license, while the model itself is available under DeepSeek's licensing terms. Enterprises can also explore its capabilities through DeepSeek Chat, a ChatGPT-like platform, or integrate it into their systems via API for commercial use.

PRESENTED BY HUBSPOT

Top-Rated Free CRM Software for Your Whole Business

If you’re looking for an all-in-one CRM platform that meets the needs of small businesses, look no further than HubSpot. With its intuitive user experience, powerful features, and scalability, HubSpot is the best CRM choice for small and scaling businesses alike.

HubSpot offers a comprehensive set of tools that can help you better understand your customers, manage leads, stay on top of customer conversations, track sales performance, and close deals faster. The connected platform enables you to grow your business faster by focusing on what matters most: your customers.

HubSpot’s CRM is 100% free to use (seriously). For free, you can store and manage up to 1,000,000 contacts with no limits on users or customer data.

Whether you’re just starting out or already scaling up, give HubSpot a try today!

Get your free CRM

🎁 Today's Trivia - Vote, Learn & Win!

The last trivia poll of this year - Answer the question below for a chance to win a 3-month membership at AI Fire Academy (500+ AI Workflows, AI Tutorials, AI Case Studies).

TODAY IN AI

AI HIGHLIGHTS

💡 AGI will be reached when OpenAI creates systems that can make close to $100 billion in total profits. After the o3 announcement, while excitement about AGI was high, Microsoft reminded everyone of their clear definition in the agreement.

📉 OpenAI’s ChatGPT, Sora, and APIs are down for many users due to an outage linked to a provider issue (the second time). The problem started around 1:30 p.m. ET, with users encountering errors or failed prompts. OpenAI confirmed “high error rates” but hasn’t provided details about the affected provider or a timeline for a fix.

💳 OpenAI is on the verge of becoming the Visa of artificial intelligence, dominating the market through exclusivity, lobbying, and regulatory influence as it transitions from innovation to monopoly-building.

📈 Goldman Sachs' 2025 AI forecasts - exciting changes are on the horizon!

DAILY AI FUNDRAISING

Bluenote raised $10 million to advance its AI platform for life sciences, streamlining regulatory workflows and reducing manual compliance work. The platform already supports 15+ applications, accelerating breakthroughs to patients.

AI SOURCES FROM AI FIRE

NEW EMPOWERED AI TOOLS

1. ✅ EasyLang AI helps you learn any language with personalized, fun, and simple AI tools.

2. 🌏 Text to Santa Videos by Gan.AI allows you to send a message and receive a personalized video from Santa.

3. 📍 Steer enables you to quickly improve your writing with an intelligent, lightning-fast assistant for any application.

4. 🛠️ Trickle lets you easily create stunning AI-powered apps, websites, and forms.

5. 📇 Revv Invest helps you discover key factors driving stock performance and invest confidently using an AI-powered search platform.

AI QUICK HITS

1. 🤖 10 Things You Should Never Say to AI Bots (Link)

2. 🌟 Agentic AI The Next Big Buzzword Everyone Will Discuss (Link)

3. 🧠 Gemini 2.0 Brings 2D Spatial Understanding to a New Era of AI (Link)

4. 🎬 New 2025 Laws Target AI in Movies, Guns (Link)

5. ⚡️ Top 10 Most Powerful Supercomputers Driving AI (Link)

AI CHART

3 key pricing models shaping the future of AI Agents

1. Platform + Hire an Agent:

• Combines a base platform fee with a usage fee for "hiring" AI agents, similar to employing a full-time worker.

• Example: Cognition’s Devin charges per AI software developer, with fees for exceeding work unit limits.

2. Platform + Outcome-Based Pricing:

• A hybrid model that includes a platform fee plus charges based on specific, measurable results.

• Example: An AI agent for call centers charges for successfully onboarding customers through outbound calls.

3. Pure Outcome-Based Pricing:

• Payment is strictly tied to achieving predefined outcomes, ideal for high-volume, results-driven use cases like sales or content generation.

What is the Best Practices for Pricing

• Start with limited usage in existing plans with a slightly higher platform fee to reduce complexity.

• Gradually shift to usage-based or outcome-based models after demonstrating value.

• Invest in infrastructure for tracking outcomes and integrating metrics into billing systems.

• Engage P&L leaders early to align with enterprise KPIs and ensure pricing strategies support business goals.

AI JOBS

• Google: Software Engineer, AI/ML/Full-stack, Google Public Sector (Link)

• Lorven Technologies Inc.: Data Science Machine Learning AI (Link)

• Hewlett Packard Enterprise: HPC-AI GTM Analyst (Link)

• Apple: Artificial Intelligence and Natural Language Processing Expert (Link)

We read your emails, comments, and poll replies daily

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