Skip to content
Models · Nov 5, 2026

Open model inference startups report exponential token-volume growth, per Interconnects analysis

Fireworks reports serving more than 40 trillion tokens per day to customized models, with open inference startups’ token volume growing near-exponentially before a recent uptick tied to GLM-4.5.

Trust71
HypeSome hype

1 source · single source

ShareXLinkedInEmail
TL;DR
  • Open model inference startups report near-exponential growth in token volume, with Fireworks disclosing over 40 trillion tokens per day to customized models.
  • The growth predates a recent uptick associated with GLM-4.5, according to Interconnects’ analysis of public disclosures.
  • Interconnects’ dashboard aggregates open model adoption and performance metrics, but the author notes measurements remain imprecise due to noisy platform metrics and marketing-related data.
  • The analysis focuses on startups such as Baseten, Fireworks, and Together AI, alongside OpenRouter’s reported token volume, while hyperscalers’ usage by model category remains undisclosed.

Interconnects founder Nathan Lambert reports that open model inference startups are experiencing near-exponential growth in token volume, based on public disclosures and the Interconnects Dashboard, an aggregator of open model adoption and performance metrics. The analysis highlights Fireworks’ disclosure that it is serving more than 40 trillion tokens per day to customers’ customized models. Lambert notes this figure represents a vertical that grew over twofold since this summer, implying total token volume for Fireworks likely exceeded 100 trillion tokens per day in the fall.

The growth trend predates a recent uptick Lambert attributes to GLM-4.5’s step change in open agents, according to a slide from a talk on the open model economy. The analysis centers on startups such as Baseten, Fireworks, and Together AI, as well as OpenRouter’s reported token volume via its API. Lambert acknowledges that measurements of open model adoption remain imprecise due to reliance on noisy platform metrics, marketing-related disclosures, and imperfect public evaluations.

Lambert’s work attempts to ground open model token volumes against closed model or hyperscaler volumes, noting that Google releases platform-wide token volumes in earnings reports. However, hyperscalers like Amazon Bedrock and Azure have not broken down usage by model category, limiting direct comparisons. The Interconnects Dashboard remains freely available, while detailed analysis is provided in paid research notes.

Sources
  1. 01Interconnects — Nathan Lambert — How big is the open model economy today?
Also on Models

Stories may contain errors. Dispatch is assembled with AI assistance and curated by human editors; despite the trust-score filter, mistakes happen. We correct publicly — every article links to its revision history. Nothing here is financial, legal, or medical advice. Verify before relying on any claim.

© 2026 Dispatch. No ads. No sponsorships. No paid placement. Reader-supported via Ko-fi.

Built by a person who cares about honest AI news.