Open-Source Models to Capture More Token Share

ollama · x · 2026-07-11

Ollama believes that the majority of tokens in the future will originate from open-source and open-weight models. A key insight quoted is that **compute scarcity** is currently masking the true economic structure of AI; while open-weight models are highly capable, running them reliably and cost-effectively remains complex. They further note that leading labs currently hold pricing power because they bundle "model + compute + reliability + access" together. However, if compute becomes more abundant and open-source toolchains become easier to use, many routine AI workloads will naturally migrate to cheaper open models.

Original post →

More from Infra

Infra channel →