Last booth day for FocoosAI at CVPR! Our CEO and CTO will be happy to introduce you our platform and code! . See you there!. Ps. We have amazing t-shirts! Come before they end!

🔥🔥.

Introducing Grok Imagine.

RT @fcdl94: We released open source our core training and models code. We’re in a path where computer vision models will require hours, no….

We’ll be in Nashville in less than a week for #CVPR2025! .@gabrosi3 will show you whether is better to show (visual prompting) or tell (open vocabulary) to achieve better performance in segmentation!.

How do you reduce your costs when using LLMs? Let us hear thoughts!. For reading the full article or subscribing to our (new) substack:

Do you really need the biggest model, long prompts, and verbose outputs? 🤷‍♂️. Smart choices = lower costs, higher efficiency, and reduced environmental impact. Let’s do more with less. 💡🔋 #AI #LLMs #Sustainability.

AI’s growth = rising energy footprint 🌍. How can we optimize LLM usage?.✅ Use smaller, efficient models (LLaMA 8B, Phi-4 16B, Qwen 7B). ✅ Minimize input tokens (prompt engineering matters!). ✅ Limit output length (shorter responses = lower cost) - just ask LLM to do it.

The cost in energy? ⚡.A medium-sized model (65B params) consumes ~3 Joules per token. - 200 output tokens ≈ powering a 10W LED for a minute 💡.- Millions of tokens/hour = megawatts of energy.- ChatGPT’s ops cost millions per month (excluding training!).

Sounds fast, right? Not quite. Each token requires a full forward pass through the model, meaning:.- Bigger models = More computation.- Longer inputs = Quadratic increase in cost.- Longer outputs = More forward passes in the model (aka more computation).

When you ask an LLM a question, it doesn’t “know” the answer. Instead, it:.1. Breaks input into smaller pieces (tokens). 2. A Transformer model computes relationships between words. 3. The model predicts and generates one token at a time.

The Hidden Cost of LLMs: What Happens Under the Hood? A Thread 🧵. You’ve probably used an genAI assistant today, certainly powered by LLMs. But have you ever wondered what does it take to answer you? 🤔. It's not just computation—it’s energy and money. Let’s break it down