Analysis of 18 AI benchmarks shows open-source models are closing the gap with proprietary ones, but at vastly different rates depending on which skill is measured.

What happened

A researcher analyzed how quickly open-source large language models (AI systems that understand and generate text) are catching up to closed-source rivals across 18 different performance benchmarks. On the Artificial Analysis Intelligence Index alone, the gap appeared to shrink to zero by December 3rd 2026, but when the same method was applied across all 18 benchmarks, the average gap remained almost flat at just under 5 months over the entire period.

What to watch

The coding index is the standout area where open-source is rapidly narrowing its disadvantage. The full set of 18 benchmark frontier plots is available for review at the bottom of the original analysis, allowing readers to assess progress across specific skill areas rather than relying on a single headline metric.