AI21 Labs

Today at sundown marks the beginning of the Jewish New Year, Rosh Hashanah. It is one of the most important holidays in the Jewish religion and throughout Israel, so we wanted to take this opportunity to wish everyone, no matter your background or religion, a Shana Tova (Happy New Year!) As we venture into the new year, our thoughts are with the families most affected by the ongoing conflicts in Israel - those that have been forced from their homes in the North & the South, our brave soldiers (many of whom are also AI21 employees), and the 101 hostages still being held in Gaza. Praying for peace, quiet and unity in the upcoming year. Here's to a meaningful, productive, and fulfilling year ahead for all. 

Breaking news — #Jamba 1.5 Mini is now available on the Databricks Marketplace! With an expanded 256K context window, Jamba 1.5 Mini offers greater context understanding, enabling deeper insights and more precise responses for your most demanding data tasks. Whether automating data analysis or supercharging complex queries, Jamba's powerful language capabilities elevate your #Databricks workflows, turning data into actionable insights with unprecedented ease. Explore our listing today! https://lnkd.in/efssxQ5Z #DataProviders #DataSharing Databricks

Last week the entire AI21 Labs team took a well-deserved break from our daily routine to close out another year (fun cultural fact: in the Jewish calendar the year starts and ends in the Fall during the Rosh Hashanah high holidays). We were able to pause, unwind, and relax together as a team, while reflecting on the past year of both successes and challenges. We decided to channel our energy into something meaningful by creating skateboards and bicycles for the children who were impacted by the October 7th attacks. Our operations team has put serving the communities of the North and South of Israel at the forefront of everything we've done this year - every event from our recent team day to happy hours features items crafted or cooked by those most affected by the ongoing war in Israel. We're proud to support these communities and grateful for the chance to come together as a company and give back.

Excited to be partnering with Databricks on AI Model Sharing! Read more below 👇

Most enterprises are still struggling to bring complex AI workflows into production. But why? Our Co-Founder & Co-CEO Ori Goshen spoke at the Calcalist כלכליסט AI conference about the primary issues enterprises are grappling with today when it comes to getting AI into production: reliability, transparency, efficiency, and data security. Ori emphasized that in order to tackle these challenges, both companies like AI21 Labs and the enterprises they serve need to think differently. With our recent release of #Jamba 1.5, we bucked the trend, introducing a model built upon a novel #LLM architecture. And we're also thinking differently in the long term - LLMs aren't the be all and end all of #AI innovation. It’s not about the biggest model—it's about the smartest, most integrated system that supports enterprises from end-to-end. Read more about Ori's talk here: https://lnkd.in/ez-2B6tH

With the #Jamba 1.5 Model Family now available on Snowflake #CortexAI, getting your AI applications & workflows into production has never been easier. Snowflake's Cameron Wasilewsky and AI21 Labs' Ian Cox, MSDS discuss the key challenges and best practices to go from prototype to production in our joint blog. Read more here: https://lnkd.in/eHT883C2

Go long (on context) with #AWS & AI21 Labs. ☁️🏈💡 https://go.aws/3BfFwyE Introducing Jamba 1.5 Mini & Jamba 1.5 Large, now in #AmazonBedrock. With context windows of 256k tokens, both models offer accessible, powerful, user-friendly platforms ideal for data-heavy tasks. Paired with low latency, they’re a touchdown for enterprises. #generativeAI

New! AI21's Jamba 1.5 model family (with fast speed, huge context window, and a super interesting hybrid architecture), is now available on Bedrock. Fire it up! Today, we are announcing the availability of AI21 Labs’ powerful new Jamba 1.5 family of LLMs in Bedrock. These models represent a significant advancement in long-context language capabilities, delivering speed, efficiency, and performance across a wide range of applications, with a 256K token context window, structured JSON output, and function calling. Jamba's key innovation lies in its hybrid design. It blends Transformer elements - the backbone of most recent AI language models - with a newer approach called "Mamba", a Structured State Space model. This combination allows Jamba to process information more efficiently, especially when dealing with long pieces of text. AI21 found that the Mamba component, while efficient, struggled with some tasks that require understanding context from examples. However, when combined with the Transformer technology, these issues were resolved, creating a system that has the best of both worlds. Built at large scale, Jamba uses a 'mixture of experts' (MoE) to provide a high throughput and small memory footprint compared to vanilla Transformers, while delivering state-of-the-art performance on standard language model benchmarks and long-context evaluations. You can use two new Jamba 1.5 models from AI21 in Bedrock: 🔷 Jamba 1.5 Large excels at complex reasoning tasks across all prompt lengths, making it ideal for applications that require high quality outputs on both long and short inputs. 🔹 Jamba 1.5 Mini is optimized for low-latency processing of long prompts, enabling fast analysis of lengthy documents and data. Fire them up, and prompt happy!

🎉 Exciting News - the Jamba 1.5 Model Family is now available on Amazon Bedrock! ⚡ Built for #longcontext tasks like content generation and summarization, #Jamba 1.5 provides exceptional speed, performance, and efficiency. With Amazon Bedrock, it’s never been easier for developers and enterprises to integrate these high-quality models seamlessly into their workflows and scale as needed. ⚙ Want to get started? Learn more about #Jamba 1.5 on Amazon Bedrock including how to access them here: https://lnkd.in/eTWN7xt2 #AWSBedrock #Amazon #AWS #AI21Labs #Jamba1.5 #LLM #AI #CloudTech AWS Developers AWS AI Amazon Web Services (AWS)

Yesterday we hosted our inaugural AI21 Labs x NVIDIA DevCon at our HQ in Tel Aviv! Over 115 enterprise AI developers & data scientists joined us in person from companies like Shopify, Intuit, Playtika, Samsung, Vimeo, AWS, Pinecone, WalkMe, Deloitte, PWC, Google, PayPal & many more. It was a full day event featuring a joint keynote with NVIDIA, followed by niche breakout sessions on topics like mastering enterprise #RAG, how & why to finetune a #LLM, the road to production with LLMs, human evaluation, configuring & developing #agents, deploying OS LLMs, benefits of #Jamba architecture, and inference/quantization. We wrapped up the day with a developer challenge focused on configuring the top benchmark for a finance data set using #Jamba (on AI21 Studio or NVIDIA NIM). Congrats to our #1 challenge winner Shirli Di-Castro Shashua!