Gartner forecasts that open-weight models will handle 90% of enterprise agent workflow steps by 2030, according to analysts Mark Petty and Raymond Paquet. These models can perform classification, tool calling, data validation and reconciliation at a quality comparable to closed models, with real-world savings of five to six times. Companies are expected to use specialised, fine-tuned models for stable, high-frequency tasks while reserving frontier models for the most complex steps. Value would therefore shift toward orchestration, which determines which model handles each step, while self-hosting would give companies greater control over costs and vendor dependencies.
Source: Read the original article


