A Mozilla report reveals the performance gap between US frontier AI models and Chinese open-weights models has narrowed to just 4.4 months, with top open models now costing only 30% of closed alternatives. The report recommends organizations default to open models for most workloads, reserving premium closed models for specialized tasks like expert professional work, high-intensity retrieval, and long context applications.
Background
Mozilla has been publishing annual State of Open Source AI reports tracking the competitiveness of open-source versus proprietary AI models. Chinese companies like Moonshot AI have rapidly closed the performance gap with US frontier developers in 2025-2026.
- Source
- Ars Technica
- Published
- Sep 15, 2026 at 08:00 PM
- Score
- 7.0 / 10