- Introducing Kaizen - AI Coding Agent Early Access
Kaizen is now available for early access — a terminal-first AI coding agent with persistent cross-session memory, multi-agent orchestration (scout, cody, and sage agents), and CAS-based undo so you can code with confidence.
- Olla v0.0.24: Agentic Workload Bugfixes
Olla v0.0.24 is a targeted bugfix release for agentic workloads, resolving issues with translator mode, agent tooling, Anthropic tool calls, and output configuration — plus improved logging throughout.
- Olla v0.0.23: Docker Model Runner, vLLM-MLX & Anthropic Passthrough
Olla v0.0.23 is a major release adding two new backends — Docker Model Runner and vLLM-MLX — plus Anthropic Passthrough support, sensible lean-config defaults, proxy path bugfixes, and expanded integration tests.
- Olla v0.0.22: Model URL Resolution & Profile Path Fixes
Olla v0.0.22 corrects model_url resolution from endpoint configuration, adds an alternative method for resolving profile paths, and ships maintenance fixes with dependency updates.
- Olla v0.0.21: Path Preservation & OpenAI Routing Fix
Olla v0.0.21 introduces a new preserve_path setting for endpoints, making Docker Model Runner integration seamless, and fixes OpenAI-compatible routing. Refreshed OpenAI profiles are also included.
- Olla v0.0.20: LlamaCpp Integration & Anthropic API Translation
Olla v0.0.20 brings back native LlamaCpp integration and introduces experimental support for Anthropic message translation, making it easier to run local models and route Anthropic-style requests through your own infrastructure.
- TensorFoundry at NVIDIA AI Days Sydney 2025
Join us at NVIDIA AI Days Sydney 2025 to see how NVIDIA's technology enables the latest AI breakthroughs, connect with peers and experts and help create what's next.
- TensorFoundry Launches: Deploy LLMs on Your Own Infrastructure
TensorFoundry officially launches its website and introduces its product suite - Olla, FoundryOS, and AgentOS - built around a simple mission: run large language models on your own infrastructure, without cloud lock-in.