How big is the open model economy today?
- ID
- 43028
- Status
- summarized
- Published
- 05 Nov 2026, 12:35 AM
- Fetched
- 05 Nov 2026, 4:13 AM
- Provider
- Interconnects
- Category
- research-analysis
- Original URL
- https://www.interconnects.ai/p/how-well-can-we-measure-open-model
- Source URL
- https://www.interconnects.ai/feed
Summary
- Score
- 6.0
- Created
- 05 Nov 2026, 4:14 AM
- Tags
- Audience
- developersai_ml_learnersai_agent_userssaas_founders
What happened
Nathan Lambert's paid Interconnects research note asks how large open-model inference is, using Baseten, Fireworks, Together AI, and OpenRouter token volume as bellwethers while noting that hyperscaler usage via Amazon Bedrock and Azure model garden is not broken out. It cites Fireworks saying at its first company conference that it serves more than 40T tokens/day to customers' customized models, a vertical up over 2x since summer, with Lambert expecting Fireworks' total to exceed 100T tokens at the next update. The note also flags GLM 5.2 as a step change for open agents and says the free Interconnects Dashboard remains available.
Why it matters
If you are comparing open-model inference providers or sizing the market, do not treat Fireworks' >40T tokens/day as total open-model share: the article says that figure covers only customized models and that Bedrock/Azure usage is undisclosed. Ask vendors for token-volume definitions and model-category splits before making pricing, capacity, or startup-opportunity bets.
Discussion angle
What token-volume metric would actually let a builder compare open-model providers—should Fireworks, Together, and Baseten disclose total versus customized-model tokens, and how should OpenRouter's open-model share be weighted?