Pulse — Poolside
poolside.ai · 496 words · saved by 1 readers
The latest data on our models usage.
20.3 trillion tokens processed across all surfaces since our first launch. All models XS.2 XS 2.1 S 2.1 M.1 Via OpenRouter Other channels Share of daily tokens (input + output) served across all channels, from our internal serving data. OpenCode, Kilo Code, Hermes, and Cline run on dedicated keys but their usage is publicly attributed to OpenRouter — together with the rest of OpenRouter they make up our total OpenRouter traffic, and each appears from the day it moved to a dedicated key — largely 21 July 2026 — with earlier usage part of Rest of OpenRouter, not new demand. Direct API…
saved by
related reading
- laguna-m1-xs2-technical-report.pdfpoolside.ai
- State of AI 2025: 100T Token LLM Usage Study | OpenRouteropenrouter.ai
- frontier model training methodologies | Alex Wa's Blogdjdumpling.github.io
- Compare AI Models: Pricing, Context & Benchmarks | OpenRouteropenrouter.ai
- Paper AI Tigersgleech.org
- How to Deploy Your Modelhtdym.sailresearch.com
- Opus 4.7's New Tokenizer: What It Actually Costs — OpenRouter Blogopenrouter.ai
- GLM-5.2: Built for Long-Horizon Tasksz.ai
- LLM Inference Performance Engineering: Best Practices | Databricks Blogdatabricks.com
- Keep the Tokens Flowing: Lessons from 16 Open-Source RL Librarieshuggingface.co
- Training Imperativesdan.io
- Hacker Trends: 18 years of Hacker News, chartedhackernewstrends.com