platoseedplatoseed
Models and infra for coding agents
Relace makes it easy to deploy production-ready coding agents. Our models are co-optimized with infrastructure to achieve SoTA performance: 10k+ token/s code merging and retrieval across million-line repositories in seconds.
Relace provides purpose-built AI models and infrastructure for coding agents, focusing on fast code retrieval, merging, and autonomous workflows to accelerate development and reduce errors.
Relace offers in-house models optimized for coding workflows, including fast codebase retrieval and a universal code merging model that edits files at 10,000 tokens per second. The system includes source control designed for agents with lightweight push/pull, automatic indexing for two-stage retrieval, and high-throughput rate limits. It supports self-hosted, VPC-isolated, and hosted deployments, enabling on-premise use while maintaining encryption in transit and at rest. Users can experiment via a hosted API, with a free tier, and pricing depends on token usage across multiple models such as Jacq, relace-apply, relace-search, relace-rank, and relace-embed.
Who it’s for: engineering teams and organizations building autonomous coding agents, CI/CD workflows, and developer tooling that require fast retrieval, merging, and reliable code generation.
Pricing page and hosted API availability; free tier and onboarding notes; mentions of enterprise/self-hosted deployments and SOC 2 compliance.
Physics -> ML
Former PhD student in machine learning at UChicago, with a math degree from Caltech.
Kicking off with Instant Apply, Code Reranker, and Embeddings
Relace releases models for AI codegen workflows, including Instant Apply for merging code snippets at high speed and an Embeddings + Reranker pipeline to locate relevant context in large codebases, aiming to reduce latency and token costs. The products target AI code generation startups and developers needing robust, efficient context management and code merging.
Formerly “Squack”
