The Information Machine
ConcludedRan 1 dayin the edition since 13 Nov 20267 sourcesupdated 13 Nov 2026

Open-source model inference volume growth

The gist

Latest: Open Model Inference Volumes Reach Tens of Trillions Daily Across Platforms

Agent workloads are consuming far more compute per user than chatbots, with DeepSeek Harness users averaging roughly 60 million tokens each per day. The scale of open model adoption on third-party inference platforms is now measurable in the tens of trillions of tokens daily, though hyperscaler footprints on Amazon Bedrock and Azure remain entirely undisclosed.

When stories like this move, I write about them in an email each morning.

One free email each morning, around 8 AM US Eastern.

The full picture

DeepSeek's open-source agent app Harness reached 1.54 million daily API users by November 2026, with those users collectively consuming roughly 92 trillion tokens per day, about 60 million tokens each. Fireworks AI was processing more than 40 trillion tokens per day as of mid-July 2026, over four times its October 2025 rate of 10 trillion, with Nathan Lambert's November 2026 analysis confirming volume remained above that level with more than 2x growth since summer. OpenRouter's weekly volume grew from 5 trillion to 25 trillion tokens in six months and was on pace to exceed 1 quadrillion tokens over 2026. An empirical study published by OpenRouter and Andreessen Horowitz in December 2025, drawing on more than 100 trillion tokens, found prompt lengths quadrupled and Chinese open-source models rose from under 2 percent weekly share in late 2024 to nearly 30 percent at peak weeks. Ollama CEO Jeffrey Morgan said in September 2026 that open models handle 80 to 90 percent of enterprise AI tokens, attributing Ollama Cloud's 150x token growth in 2026 to a coding-agent surge early in the year and a second wave in April following the OpenClaw and Hermes agent projects.

How it developed
First in the edition · 13 November 2026
8 November 2026in the 13 Nov edition

OpenRouter and Andreessen Horowitz published two empirical studies in December 2025 on more than 100 trillion tokens, finding prompt lengths quadrupled to over 6,000 tokens and Chinese open-source models rose from under 2 percent to nearly 30 percent weekly share at peak.

By November 2026, DeepSeek's agent app Harness reached 1.54 million daily users consuming 92 trillion tokens per day, while Fireworks AI exceeded 40 trillion daily and Ollama's CEO said open models handle 80 to 90 percent of enterprise AI tokens.

4 November 2026

Lambert's Interconnects analysis confirms Fireworks above 40T tokens/day, more than 2x since summer; notes measurement imprecision across open model platforms

4 September 2026

Ollama CEO Jeffrey Morgan says open models handle 80-90% of enterprise AI tokens; Ollama Cloud up 150x in 2026

6 April 2026

Fireworks AI CEO Lin Qiao reports 15 trillion tokens per day, up from 13 trillion, and forecasts exponential growth

4 December 2025

OpenRouter/a16z empirical study published, covering 100T+ tokens; platform processing 1T tokens/day; Chinese OSS models near 30% peak weekly share

Sources
2 more sources

Get the daily email

When stories like this move, I write about them in an email each morning.

One free email each morning, around 8 AM US Eastern.

See the latest edition · Privacy