Company overview
Gamma is an AI-powered platform that enables users to quickly generate creative content, including presentations, websites, and social media content, without any coding or design experience.
Gamma is a product-led growth company: its massive user growth has been driven by fast, high-quality self-serve experiences.
Inference challenges
Speed defines Gamma’s product experience. Gamma aims to produce results at the speed of thought, and the time it takes to generate an image or presentation is time the end user spends waiting.
When Gamma first launched in 2023, it was largely built around closed-source image generation models. But as its user base grew and the quality of open-source image generation models improved, Gamma was interested in leveraging open-source models for a few reasons:
Higher performance: Closed-source models were slow. Meanwhile, open-source models let you control your entire image generation pipeline so you can optimize it for the lowest possible latency and highest throughput.
Lower costs: With millions of platform users, open-source models would help Gamma serve billions of image generations more cost-efficiently at scale.
Gamma wanted to generate entire presentations in under 10 seconds – the time it took some closed-source models to produce a single high-resolution image. That said, Gamma had no internal ML team and it didn’t want one. It was looking for a partner to deploy open-source models at scale without scaling its internal engineering teams – that’s when the Gamma team reached out to Baseten.
Gamma is a lean team: we try to only focus on the things that set us apart and partner for everything else. We work with Baseten so we can provide the fastest product experience with open-source models, while we focus our expertise on delivering the best AI-powered design platform.
Solutions
Gamma first met with Baseten’s forward deployed engineers (FDEs) to test out different image generation models for both quality and speed. Quality is paramount for the Gamma team – new open-source models need to meet their high bar before being integrated into the platform.
From SDXL to Flux and Qwen, Baseten’s FDEs worked as an extension of Gamma’s engineering team to optimize the performance of new model releases within hours or days. The models Gamma adopted were further optimized for ultra-low latency and higher cost-efficiency, enabling fewer replicas to be used without noticeably affecting performance.
Speed is critical for Gamma. We’re a PLG company: the faster we can deliver something great to our users, the happier they are with the product. That’s why we partner with Baseten to serve our open-source image generation models. We generate millions of images a day on Baseten for our 70+ million users with ultra-low latency and high throughput.
Results
Gamma quickly offloaded its AI infrastructure and model management to Baseten’s engineers. With Baseten, Gamma produces 3+ million images per day quickly, reliably, and cost-efficiently.
Gamma is famous for producing a huge amount of product with a small team – so much so that The New York Times wrote an article about it. With Baseten, the Gamma team has scaled to over 70 million users without a machine learning team or a single infrastructure hire. The Gamma team can count on Baseten for benchmarking and testing new models, and then optimizing their performance for production.
By partnering with Baseten to power its open-source image generation models, Gamma achieved:
30%-80% faster image generation per model
20% improved efficiency (reduced replica count without noticeably affecting latency)
70+ million users and $100+ million ARR with a lean 40-person team
International scale for its global customer base
Billions of images generated (~3M per day)
Baseten’s FDE team has effectively been our team of in-house ML inference specialists. By partnering with Baseten, we’ve been able to scale to over 70 million users and billions of requests. We’ve never had a need to scale our AI or infrastructure teams, and we haven’t made a single hire for either.
What’s next
Baseten has been partnering with Gamma since its first launch in 2023. Now, Baseten and Gamma partner to create text, visuals, and images for AI-generated presentations.
Looking ahead, Baseten will continue to be a reliable partner for Gamma to test the latest and greatest open-source image generation models. When a new model drops, Baseten’s FDEs set it up via an API endpoint for Gamma to test in their existing workloads, with rapid iterations on model performance and the uptime required for mission-critical workloads.
Gamma is free to use: check out gamma.app to start designing.