Open-weight AI reportedly trails frontier models by about four months
A post summarizing a Mozilla report describes uneven economics: open models handled roughly 20% of measured OpenRouter usage but captured about 4% of model-layer revenue in a cited 2025 window. The revenue figures predate the report’s 2026 usage data.
TLDR
A post summarizing Mozilla’s 91-page report says open-weight AI is about four months behind frontier models. Eight of OpenRouter’s 10 most-used models by August token volume were open-weight. The post says Mozilla attributes much of the usage–revenue mismatch to pricing, with closed models costing roughly six times more per call at about 90% capability parity. It also relays Mozilla’s warning that the cited 2025 revenue figures are older than its 2026 usage data, so the revenue split may have changed. A follow-up post from the same author says DeepSeek, Moonshot and Z.ai have raised about $21 billion combined, arguing that open-weight AI is no longer a capital-light alternative.
Combined views
5.9K
2 Sources, first seen 23h ago
Open-weight AI reportedly trails frontier models by about four months
A post summarizing a Mozilla report describes uneven economics: open models handled roughly 20% of measured OpenRouter usage but captured about 4% of model-layer revenue in a cited 2025 window. The revenue figures predate the report’s 2026 usage data.