Dipayan Mukhopadhyay’s Post

https://lnkd.in/ewqqQQrr Fascinating explainer on how to build a GPT by Andrej Karpathy (OpenAI). Very technical, but you get a faint idea on how GPT's work...

GPT-4o can harmonize with itself.

OpenAI's newest model GPT-o1 is out, and it's INSANE! It can beat PhD's and solve IMO and competitive coding problems with ease. Full review video: https://lnkd.in/evT5CmV6

GPT-4o. It’s smarter, in most ways, cheaper, faster, better at coding, multi-modal in and out. It’s Gpt-4 Omni.

While #LLM models are amazing tools, inference and especially training have huge costs (both monetary 💰 and environmental 🌍 ). We already had platforms such as https://huggingface.co to share model weights, lowering/eliminating the training expenses. But now, it's great to see that big players are also starting to optimize inference cost. The new OpenAI #GPT-4o model is an amazing step in the right direction! 👏

GPT 4o is here!

Groq Mistral is fast and impressive. Below article gives a good breakdown on performance and quality to GPT-4. Check out Groq Labs https://lnkd.in/gaE9ht5m https://lnkd.in/g5pxTRNE

Watching OpenAI's Spring Update this morning (with some entertaining insight from my daughter) and launch of GPT-4o was yet another reminder, that we're truly in a golden age of change. The capabilities and potential continue to expand and offer a fundamental shift in how we adopt and interact with technology. As technologists, we need to continue enabling our peers, leaders, families and friends to safely experiment, appreciate and drive value from these advancements, lest they be left behind. Certainly GPT-4o appears to lift the bar again on performance (50% cheaper and faster??). A few quick GPT-4o headline features: 📳 Text, Audio, Image and Video input and output 💱 Improved tokenization, lowering token usage and costs. 🚤 Significantly lower latency and improved performance on audio https://lnkd.in/gTj9BP9x Also Falcon 2 11B release to read through and consider. Great start to a week.

GPT-4o mini is OpenAI’s latest creation, designed to be a smaller and more budget-friendly version of the powerful GPT-4 model. https://lnkd.in/dmmXewMY - GPT-4o mini is an affordable and compact version of the GPT-4 model. - It supports both text and image inputs, providing text outputs. - GPT-4o mini scores 82% on the MMLU benchmark, surpassing other small models. - The model is more than 60% cheaper than GPT-3.5 Turbo. - It is designed to be cost-effective without sacrificing performance.

The GPT-4o demo is amazing! 💯 Watch this video that OpenAI shared less than 24 hours ago if you aren't one of the 2 million viewers. 😂 Text, voice, and reasoning are combined in GPT-4o. It's impressive to see: 🔸It's okay to cut him off (even if it's not very polite), 🔸Ask him to solve an equation based on an image capture, 🔸To get him to tell you a story, ask him to use different voices, 🔸Face capture can be used to analyze emotions. 💥 With a 50% lower price and 2x faster response time than GPT-4 turbo. What do you think of this demo? https://lnkd.in/eTfgtWW4 #GPT4o #GPT4 #OpenAI