“Through the partnership with Baseten, we have the ability to serve different types of models, which means that my team can really focus on the ones that are the best fit.”
Please rotate your device
To continue exploring this website, please rotate your device to a vertical orientation
Customer obsessed
Trusted for your most important workloads.
$1.9M
saved via model training
2-3x
lower latency with AI meeting notes
3x
cost-savings vs. closed-source
80%
lower latency per image
Featured customer stories
The teams building on Baseten
“Baseten gives Canva the confidence to launch AI features to hundreds of millions of people, knowing that the capacity will be there when we need it.”
“Our customers really care about speed, reliability and the ability to deploy custom fine tuned models for financial services. Baseten met all three criteria far and above all the competition.”
“Baseten lets us productionize our models in days, not months.”
“We generate millions of images a day on Baseten for our 70+ million users with ultra-low latency and high throughput.”
“Baseten Loops helps us test new ideas in multi-turn agentic RL with on-demand capacity and a flexible training SDK. We can experiment with open-weight models while keeping control over our training logic.”
All customer stories
Powered by Baseten
OpenEvidence delivers instant, accurate medical information with the Baseten Inference Stack
OpenEvidence partners with Baseten for their inference infrastructure to focus on what they do best: making exceptional tools for physicians.
- lower latency
- 78%
- faster deployment processes
- 6x
How Writer helps businesses transform with custom medical and financial LLMs
Writer, the leading full-stack generative AI platform, launched new industry-specific LLMs for medicine and finance. Using TensorRT-LLM on Baseten, they increased their tokens per second by 60%.
- parameter custom LLM
- 70B
- higher TPS
- 60%
How Zed is reimagining the code editor from the ground up
By partnering with Baseten, Zed achieved 45% lower latency, 3.6x higher throughput, and 100% uptime for their Edit Prediction feature.
- lower p90 latency
- 45%
- higher throughput
- 3.6x
Rime serves speech synthesis API with stellar uptime using Baseten
Rime AI chose Baseten to serve its custom speech synthesis generative AI model and achieved state-of-the-art p99 latencies with 100% uptime in 2024.
- milliseconds p99 latency
- <300
- distinct voices
- 200+
Wispr Flow creates fast and accurate voice dictation on Baseten
Wispr Flow runs fine-tuned models with Baseten to provide seamless dictation across every application.
- milliseconds end-to-end
- <700
- latency measured
- P99
How Gamma makes building presentations criminally fun with 5x faster image generation
With Baseten handling inference, Gamma's lean team stays focused on building a product users love.
- hires for ML or AI infra
- 0
- images generated per day
- 3+ million
How Superhuman achieves 80% faster embedding model inference with Baseten
Superhuman cut P95 latency by 80% across dozens of custom embedding models in just one week after adopting Baseten Embedding Inference.
- lower latency
- 80%
- p95 response time
- 100 ms
Latent delivers pharmaceutical search with 99.999% uptime on Baseten
Latent Health uses Baseten to power fast, reliable clinical AI.
- uptime
- 99.999%
- P90 end-to-end latency
- 600 ms
Building AI Agents, Open Code, and Open Source Coding with Dax Raad
With fast, reliable inference from Baseten, OpenCode keeps developers in their flow.
Praktika delivers ultra-low-latency transcription for global language education with Baseten
With Baseten, Praktika delivers <300 milliseconds latency empowering language learners worldwide with a seamless conversational and learning experience.
- milliseconds latency
- <300
- cost savings
- 50%
From datasets to deployed models: How Oxen helps companies train faster
Oxen AI is a platform focused on helping companies accelerate the path from raw datasets to fine-tuned, production-ready AI models. They partner with Baseten to handle the infrastructure layer while they innovate on the customer facing side.
- Lower inference time
- 50%
- Training job scheduled
- < 1 min
Scaled Cognition offers ultra-fast AI agents you can trust
Scaled Cognition cut latency 40% and hit <120 ms TTFT with Baseten’s hybrid autoscaling—powering ultra-fast, enterprise-grade agentic workflows.
- lower end-to-end latency
- 40%
- ms TTFT
- <120
Patreon saves nearly $600k/year in ML resources with Baseten
With Baseten, Patreon deployed and scaled the open-source foundation model Whisper at record speed without hiring an in-house ML infra team.
- hours dev time saved yearly
- 440+
- annual resources saved
- $600k
How Sully.ai returned 30M+ clinical minutes to healthcare using open-source models.
How Sully.ai addressed their latency, cost and quality challenges by transitioning its Inference Stack to open-source models running on Baseten.
- Inference cost savings
- 90%
- Lower median latency
- 65%
How OpenEvidence puts peer-reviewed research in the hands of nearly half of all U.S. physicians
OpenEvidence trains their custom models on Baseten Training, utilizing key features such as code customizability, on-demand compute, and one-click deploys to production inference.
- minutes to get started
- 30
- saved via model training
- $1.9M
How World Labs is building large world models, pushing the boundaries of 3D
World Labs is building the optic nerve of AI
Posit launches real-time AI code suggestions with Baseten
Posit ships context-aware AI code completions for RStudio, powered by Baseten's fast, flexible inference.
- minute to acquire compute
- 1
- faster than other providers
- 60%
How Hebbia powers AI workflows for the world's leading financial institutions
Hebbia powers real-time financial intelligence for top institutions with Baseten's low-latency inference infrastructure.
- improvement in TPS
- 2.5x
- improvement in TTFT
- 4x
How OpenCode powers the AI coding agent developers rely on every day
OpenCode partnered with Baseten to power Zen, their model gateway for coding agents, achieving 5x faster inference, sub-second TTFT, and a 33% cost reduction while scaling from 40k to 10M monthly active users.
How Baseten powered Poolside's model launch in record time
When Poolside set out to launch their Laguna model family to the world, they needed more than a hosting provider. They needed a partner who could move as fast as they did. Baseten deployed a full production inference platform for them in just 7 weeks.
- From project kickoff to production
- 7 weeks
- P50 TTFT for Laguna M.1 Model
- 605 ms
How EliseAI trains specialized models for healthcare and housing operations with Baseten Applied Research
EliseAI is the domain expert in AI-powered housing and healthcare and tackling some of the hardest operational problems in the economy. By partnering with Baseten Applied Research, EliseAI now owns the intelligence powering its product.
- reduction in inference cost
- 60%
- reduction in latency
- 80%
How Speechify makes audio the default with real-time text-to-speech
Speechify synthesizes 161B+ characters per month for 60M+ users. With Baseten, Speechify cut costs by 44%, p99 latency by 30-50%, and got 4.5x faster cold starts.
- lower cost per million characters
- 44%
- lower p99 inference latency
- 30-50%
How SpeechifyAI built an ultra-low-latency voice agent without compromising on voice quality
SpeechifyAI is the developer arm of Speechify, the world's leading voice AI platform serving more than 50 million users, with APIs for text-to-speech, voice agents, and conversational AI.
- lower latency per turn vs. a stitched multi-vendor pipeline
- 75–150 ms
- median streaming-ASR transcription / LLM time-to-first-token
- 420-480 ms
How Parallel Web Systems pioneers programmatic web with high-throughput agentic inference
Parallel powers high-throughput agentic web research with Baseten's throughput-optimized inference and training-to-deployment loop.
- higher throughput
- 3x
- improvement in latency
- 2x
How You.com cut costs by 30% with open-source models on Baseten
You.com enables you to search the web, extract clean content, and generate grounded answers for Agents and LLMs.
- To migrate
- <1h
- Cost Reduction
- 30%
How Notion achieved 2-3x lower latency with AI Meeting Notes
Notion built AI Meeting Notes with a custom, streaming transcription and diarization pipeline on Baseten and cut costs by 50%
How Magicare AI helps post-acute care grow compliantly & profitably
Magicare AI is the AI-native operating system for post-acute care, agentically automating healthcare administration and clinical services for nursing homes, behavioral health, and home health. Magicare reduces admission decisions from 30-60 minutes to <1 minute and automates authorizations, reimbursement, and compliance.
- reduction in cost
- 33%
- TPM of traffic handled
- 60M
Mirendil runs frontier RL on Baseten
Mirendil runs rollouts on Baseten's RL stack and global fleet of clusters.
- throughput improvement
- 1.9x
- delta weight syncs across global clusters
- <40s
LangChain trains custom models for LangSmith Engine with Loops
Baseten Loops provides managed infrastructure for fine-tuning models through an API, with support for supervised fine-tuning, reinforcement learning, and long-context workloads.
How Harvey works with Baseten Applied Research
New research with Harvey shows how recursive language model harnesses and post-training can improve AI performance on complex M&A diligence.
The team behind your team.
Tell us what you're building. We'll partner with you on your workload, performance goals, and path to production.
