Enables Quora’s AI-powered chat platform, Poe, to deliver multiple advanced foundation models to users
Running Anthropic’s Claude models on Vertex AI achieves 40% better performance compared to other cloud infrastructure platforms for serving Anthropic on Poe
Extends reach of gen AI models and Quora’s Poe platform to a global audience
Quora uses Vertex AI to deploy foundation models, such as Anthropic’s Claude 3.5 Sonnet, to power its AI chat platform, Poe.
Since 2010, Quora has managed a platform that enables people from all over the world to share knowledge through a question-and-answer format. While the site had always been driven by person-to-person interactions, Quora saw the chance to provide users with an entirely new experience using foundation models and other powerful gen AI technologies.
This laid the groundwork for Quora’s newest AI-powered chat platform: Poe.
“We asked ourselves what an ideal new product would be if we started it from scratch using the latest and greatest gen AI,” says Spencer Chan, Product Lead at Poe by Quora. “Since each major gen AI player was building models with different strategies, we quickly realized that we needed a multi-model approach to achieve our goals.”
Like Quora’s main site allows people to share a virtually infinite range of perspectives, the Poe platform would enable users to interact with an array of gen AI models that provide insightful answers to questions instantly. It was a novel idea to democratize access to some of the world’s best gen AI models. This led Quora to Google Cloud because Vertex AI Model Garden matched the company’s vision to integrate leading foundation models into Poe — such as Anthropic’s Claude models.
Poe now delivers bots that allow users to engage with the latest models on a variety of topics from cooking advice to help with building applications. Each bot allows users to leverage different foundation models, and multiple gen AI models can even be present in one person’s chat, allowing users to compare different perspectives and answers from different models.
Quora uses Vertex AI’s enterprise-grade infrastructure and robust tooling to go from experimenting with models to deploying applications in production quickly.
Gemini and Claude are hosted on and orchestrated by Vertex AI, which reduces the infrastructure management burden and cost on the Quora team. In addition, Vertex AI enables Quora to go live with the newest Anthropic models on the first day of their release. Because of this, Poe users will receive access to the most high-performing Claude models the day they are available.
At the same time, because of breakthroughs enabled by Google Cloud TPU v5e, Quora improved performance by 35-40% by running Claude 3.5 Sonnet on Vertex AI compared to other clouds it tested. Vertex AI also offers provisioned throughput for Claude models, providing Quora with reserved capacity.
With the constant evolution of Poe, Quora introduced a new feature called Previews that supports interactions with Poe-generated web applications. Poe users, even those with no coding experience, can create custom interactive experiences such as games, animations, and data visualizations. Previews is powered by foundation models that excel at coding activities like Claude 3.5 Sonnet and Gemini 1.5 Pro.
Poe makes Claude 3.5 Sonnet available to users globally, while ultra-low latency and efficient networking in Google Cloud deliver positive user experiences across regions. Poe users can turn to Claude for reliable, conversational, and informed answers related to the new knowledge they hope to gain.
Quora continues to find new ways to use Google Cloud to maximize performance and minimize costs across its Poe platform. For instance, the company takes advantage of provisioned throughput and dedicated machines to run heavy traffic and workload sizes.
Quora’s mission is to share and grow the world’s knowledge.
Industry: Technology
Location: United States
Products: Model Garden on Vertex AI
About Google Cloud Partner- Anthropic
Anthropic is an AI safety and research company that builds reliable, interpretable, and steerable AI systems.