Cerebras Systems

Meet Cerebras Inference – the fastest inference for generative AI! 🏎️ Speed: 1,800 tokens/sec for Llama 3.1-8B and 450 tokens/sec for Llama 3.1-70B, 20x faster than NVIDIA GPU-based hyperscale clouds. 💸 Price: Cerebras Inference offers the industry’s best price-performance at 10c per million tokens for Llama 3.1-8B and 60c per million tokens for Llama-3.1 70B. 🎯 Accuracy: Cerebras Inference uses native 16-bit weights for all models, ensuring the highest accuracy responses. 🔓 Access: Cerebras Inference is open to everyone today via chat and API access. All powered by our third-generation Wafer Scale Engine (WSE-3). Try it now 👉 https://lnkd.in/gEJJ2pfY Press Release: https://lnkd.in/gtF5fxHt Blog: https://lnkd.in/gZ46q4cD

The latest SC24 Newsletter is out and it features Cerebras, Sandia National Laboratories, Lawrence Livermore National Laboratory, and Los Alamos National Laboratory as one of the six finalists of the prestigious Gordon Bell Prize!   Our collaboration has enabled a remarkable 457x speedup in molecular dynamics simulations using Cerebras' Wafer-Scale Engine.   Read about the finalists here: https://lnkd.in/g_XUj4ZN

Behind the scenes: How llama models are powering AI News In this fireside, we chat with Shawn swyx W, the editor of the Latent Space Podcast and AI news newsletter. Swyx gives a behind the scenes on how he uses llama models to produce his content that has over 3M viewers. Swyx also chats about the recent llama 3.2 drop and what use-cases he expects to move to the edge.

In this fireside interview, Ankur Goyal, CEO of Braintrust, shares his experiences with Llama models! 1. Why tool calling is the most common use case 2. The importance of platforms that are independent of the large model labs 3. How llama models beat more established models in production

Cerebras is proud to sponsor the Conference on Language Modeling (#COLM) this week at the University of Pennsylvania. We will be presenting a poster on the Jais series of bilingual Arabic models developed with Core42 and MBZUAI. Jais outperforms state-of-the-art Arabic models, with a 7.5% improvement in Arabic MMLU for Jais 70B adapted compared to LLaMA-2 70B. These models show significant improvement on Arabic multi-turn benchmarks, paving the way for more accurate and nuanced language understanding. Have questions about LLMs and Multilingual Models? Ask us anything: www.cerebras.ai/contact-us

Congratulations to Sandia National Laboratories, Lawrence Livermore National Laboratory, and Los Alamos National Laboratory for being named finalists for the 2024 Gordon Bell Prize! Your groundbreaking 457x speedup in molecular dynamics simulations performed on Cerebras Wafer-Scale Engine paves the way for new advancements in materials research and scientific discovery. https://lnkd.in/g54fNpny

In this fireside interview, Sophia L., a current Greylock Partner and previous Scale Senior PM shares on the long-going Scale and Meta partnerships. She discusses the challenges around creating data to train large scale LLMs like llama models, and how this was fundamentally different from the data for autonomous driving models Scale previously produced.

Last week, Cerebras hosted Llamapalooza, a night all about fine-tuning and using llama models in production. Thank you to the hundreds of llama community members who joined us for the talks, panel, and after-party. Shout out to our speakers Charles Frye, Roland Gavrilescu, Shawn swyx W, Jerry Liu, Angela Yeung, Daniel Kim See you all at the next one!

Cerebras Inference: highest speed, lowest latency. Try it out! https://lnkd.in/gEJJ2pfY

Major performance update: Llama3.1-70B now runs at 560 tokens/s! 24% faster in 3 weeks Available now on Cerebras Inference API and chat https://lnkd.in/dTF-6yGP