We've partnered with Modular to create Large Scale Inference (LSI), a new OpenAI-compatible inference service. It's up to 85% cheaper than other offerings & can handle trillion-token scale. We originally created it at the request of a major AI lab to do large scale multimodal synthetic data creation at tens of millions of dollars cheaper than alternatives. Unlike other services, LSI uses market pricing. The token prices are a function of the underlying compute price on sfcompute & current system load. In other words, it's always the best price, forever. Many have claimed inference is a race to the bottom, so we sell you the bottom. LSI is built on Modular's MAX & Mammoth inference framework, a highly optimized redesign of serving stack, with support for multiple chips. Between Modular's world class engineering & SFC's ruthless price optimizations, we're committed to creating the highest performance, best priced inference in the world. To get a quote, please reach out. We run a bespoke process with each customer to optimize for your specific requirements, and only roll out when the price & performance are hitting your expectations. While we support most open source models (DeepSeek, Llama, Qwen, etc), our bespoke process lets us run custom models or niche requirements depending on your circumstances. DMs are open or you can reach out in the link below.