Tokenless
Tokenless provides an intelligent model router that cuts AI inference costs in half by automatically switching to the most cost-effective model that maintains frontier performance.
At a Glance
- AI startups
- Enterprises with high LLM spend
- Developers of agentic workflows
AI Tools by Tokenless
(1)
Tokenless
AI Inference Router for LLMs
Discussions
No discussions yet
Be the first to start a discussion about Tokenless
Latest News
Launch HN: Tokenless (YC S26) – Automatic model switching to save money
Building Tokenless: mixing models for cheap frontier performance
Introducing Tokenless: The intelligent model router that sends each request to the rightsized model
Products & Services
High-performance router that reaches SOTA on benchmarks by mixing models like GPT-5.6, Claude Opus 5, and Deepseek V4 Pro.
Cost-optimized routing mode designed to maximize savings for AI inference.
A drop-in replacement for OpenAI and Anthropic APIs that automatically routes requests to the most efficient model.
Market Position
Positions against manual 'model astrology' (heuristics) by providing research-backed, live-streaming routing decisions that factor in prompt caching and model confidence.
Leadership
Founders
Rohit Agarwal
CEO and Head of Research. PhD Candidate at Princeton University focusing on Graphics. Former Graduate Student Researcher at Princeton.
Andrew Liu
Co-Founder. Former Researcher at Google DeepMind. Graduate of University of California, Berkeley.
Kevin Wang
Co-Founder. Graduate of University of California, Berkeley (EECS, 2024).
Executive Team
Rohit Agarwal
CEO and Head of Research
PhD Candidate at Princeton, expert in graphics and AI research.
Andrew Liu
Co-Founder
Former Google DeepMind researcher with expertise in 3D computer vision and generative models.
Founding Story
Tokenless was started to address the significant waste in AI inference spend, where companies often use expensive frontier models for mundane tasks. The founders developed a fan-out routing algorithm that watches models 'think' in real-time to select the optimal model.
Business Model
Revenue Model
Usage-based with support for 'Bring Your Own Keys' (BYOK).
Pricing Tiers
$20 of free credit provided upon sign-up.
Usage-based routing focused on performance and cost balance.
Usage-based routing focused on maximum cost reduction.
Target Markets
- AI startups
- Enterprises with high LLM spend
- Developers of agentic workflows
- AI agents
- Coding agents
- Chatbots
- Copilots
- Mundane task automation (file search, summarization)