Researchers from UC Berkeley, Stanford, and Databricks just introduced GEPA, a prompt optimization method that ditches reinforcement learning. Instead of thousands of trial-and-error runs guided by simple scores, GEPA has models reflect on detailed execution logs and error traces, beating traditional methods with up to 35 times fewer trials.
TwelveLabs just locked in $30 million in fresh strategic funding, bringing its total to $107.1 million, to advance its dedicated video search and analysis AI. Backed by Snowflake, Databricks, and In-Q-Tel, the startup also hired former SK Telecom CTO and Siri architect Yoon Kim as president. Why it matters: While tech giants push general multimodal models, TwelveLabs bets purpose-built video models will dominate enterprise archives and security workflows.
Snowflake just launched Arctic LLM, an enterprise-focused open model licensed under Apache 2.0 for research and commercial use. Using a 480B-parameter MoE architecture that only activates 17B parameters at a time, it cost just $2 million to train and targets SQL generation and corporate coding workloads. Snowflake claims it outperforms rivals like Databricks’ DBRX on database tasks.
Cloudflare just launched a suite of edge tools—Workers AI, Vectorize, and AI Gateway—to help developers deploy models on nearby GPUs with pay-as-you-go pricing. It also includes vector storage and traffic caching to keep costs under control. The move could make running open models much cheaper and faster without central cloud lock-in.
NVIDIA just introduced TensorRT-LLM, an open-source library built to accelerate large language model inference on modern GPUs. It brings in-flight batching and FP8 quantization, delivering up to an 8x throughput boost on H100 GPUs and running Llama 2 4.6x faster than A100 systems. Why it matters: teams can now tap peak GPU parallelism via a modular Python API without writing low-level CUDA, slashing data center running costs.
Databricks just announced the acquisition of generative AI startup MosaicML in a massive $1.3 billion deal. Known for efficient models like MPT-7B and MPT-30B, MosaicML specializes in helping companies train and fine-tune LLMs on their own proprietary data at lower costs. Why it matters: Enterprise customers can now build secure, tailored models directly inside Databricks Lakehouse without exposing sensitive data to third parties.
