EgoCraft, Inc.

07/26/2024

Imagine feeding an entire library of knowledge into a model query—that's now possible with the latest open-source models from Meta and MistralAI, boasting 128k context windows. A sharp contrast with traditional RAG techniques, which required piecemeal inputs. Now you can include vast amounts of related data at once. And significantly cheaper than GPT-4o. 🎉

07/26/2024

Impressive! DeepMind's AI has achieved a silver medal level in solving International Mathematical Olympiad (IMO) problems. The potential applications for this are mind-blowing. Demonstrates just how far technology has come in understanding and solving complex challenges.

Breakthrough models AlphaProof and AlphaGeometry 2 solve advanced reasoning problems in mathematics

07/25/2024

Just published today by Sam Altman is an op-ed in the Washington Post on the future of AI, emphasizing its critical role in national security. He argues for U.S. leadership in AI development to ensure democratic values prevail and to prevent authoritarian regimes from exploiting AI for control. Altman stresses the importance of bipartisan support for AI advancements and the need for a global coalition led by the U.S. to shape AI's future positively. A critical read for these rapidly changing times.

We must work to ensure that a democratic vision for artificial intelligence prevails over an authoritarian one.

07/25/2024

🚀 Mistral has launched Mistral Large 2, enhancing the AI ecosystem with improved code generation, mathematics, and reasoning.

- 128k context window
- Multilingual support
- Advanced function calling capabilities

All in all, a great addition for researchers and developers. Available for research and non-commercial use, with commercial licenses upon request. Exciting times for AI advancements! 🎉

Today, we are announcing Mistral Large 2, the new generation of our flagship model. Compared to its predecessor, Mistral Large 2 is significantly more capable in code generation, mathematics, and reasoning. It also provides a much stronger multilingual support, and advanced function calling capabili...

07/23/2024

Today Meta unveiled Llama 3.1, their most advanced models to date. 🌟 Key features include:

- 128K context length
- Support across eight languages
- Llama 3.1 405B, an open-source AI model with unmatched capabilities
- New security tools like Llama Guard 3 and Prompt Guard
- Collaboration with over 25 partners like AWS and Google Cloud

Read more about Llama 3.1 405B at

Use Meta AI assistant to get things done, create AI-generated images for free, and get answers to any of your questions. Meta AI is built on Meta's latest Llama large language model and uses Emu, our...

07/19/2024

Woah! Too fast, too soon, apparently, for the EU 😲 Due to regulatory concerns, Meta will not be releasing its new multimodal Llama model in the EU in the coming months. These models handle video, audio, images, and text, and will be integrated into products like smartphones and smart glasses.

Will likely put European companies at a competitive disadvantage and increase market fragmentation, as they rely on different AI models. A larger, text-only version of Llama 3 will be available in the EU soon.

The issue stems from compliance with GDPR, not the AI Act.

The social media giant says EU regulators haven't provided clarity.

07/18/2024

For whatever reason there's been a burst of little LLMs recently 🌟

Here's a list:

Mistral NeMo: Adaptable and fast, 128k context window
GPT-4o mini: Cost-efficient, high performance, low latency
SmoLLM: 135M, 360M, and 1.7B parameter versions
Mathstral: Excels in mathematical reasoning
Codetral Mamba: Efficient in-context retrieval up to 256k tokens
H2O Danube3: Local processing on mobile devices

All pushing the boundaries of what's possible with smaller LLMs. Think Microraptors and other little dinos of the Early Cretaceous 🦖

07/18/2024

Current LMSYS rankings - GPT-4o still leads the way, with 3.5 Sonnet nipping at its heels

https://chat.lmsys.org/?leaderboard

07/18/2024

Today OpenAI introduced GPT-4o mini, their most cost-efficient small model yet. 🚀

Most interesting - currently scores 82% on MMLU, outperforming GPT-4 on LMSYS leaderboard

Sleek, small(er), adaptable, fast

Introducing the most cost-efficient small model in the market

07/18/2024

Mistral has just released NeMo 🚀 a highly adaptable, 128k context window open-source model, a successor to the popular Mistral 7B
- great for large documents
- fast input and output processing
- easy to fine-tune
- no commercial restrictions
Cheers to the growing value of open-source LLMs in the AI ecosystem 🌟

Mistral NeMo: our new best small model. A state-of-the-art 12B model with 128k context length, built in collaboration with NVIDIA, and released under the Apache 2.0 license.