New Artificial Analysis Index Widens the Gap Between Open and Closed Models
The latest Artificial Analysis intelligence index reveals a widening performance gap between open-weights and closed AI models. Proprietary systems now dominate the top rankings, while open alternatives trail further back than in previous measurements.
This assessment consolidates multiple AI benchmarks into a single intelligence score for each model. The result is a direct comparison that tracks how quickly both sides of the AI ecosystem are advancing.
What the Index Tells Us
The Artificial Analysis index evaluates models on reasoning, coding, mathematics, and general knowledge. It converts performance on established benchmarks into a composite score, allowing a clear ranking of the best available systems.
The newest data shows closed models pulling away. Commercial labs continue to invest heavily in training compute, data curation, and post-training refinement, and that investment is reflected in their scores.
The key finding: open and closed models are moving in opposite directions on the index, and the space between them keeps expanding.
Open Models Are Losing Ground
Open-weights models still produce strong results in specific categories. Some match or beat proprietary systems on coding tasks, and several stand out for efficiency at smaller scales.
But on overall intelligence scores, they trail. The latest index update shows the top positions occupied entirely by closed models, with open-weights systems spread across the lower tiers.
Key Takeaways
- Closed models take the top slots: Every leading position belongs to a proprietary system.
- Open models win niche battles: They excel in select tasks but lose the composite comparison.
- The gap is now wider: The distance between the best closed and best open models has grown since the previous update.
Why the Gap Matters
This widening divide changes the decision calculus for developers and enterprises. Open-weights models offer transparency, control, and lower cost, but raw capability is harder to justify when the performance gap keeps growing.
The open-source AI community has long argued that open models would converge with proprietary ones over time. The new index challenges that assumption.
The data suggests open development is not closing the distance. It is falling behind.
The Cost Question
Efficiency remains a bright spot for open models. Several open-weights systems deliver competitive output while requiring significantly less compute, which lowers deployment and inference costs.
That advantage, however, does not offset the capability gap for users who need maximum performance. The index separates these two dimensions, and the difference is stark.
For organizations, the tradeoff is now sharper than ever: pay for premium performance from closed providers, or accept lower benchmark scores in exchange for openness and data control.
What Comes Next
Future open-weights releases could change the trajectory. The community continues to produce ambitious models, and architectural innovations may narrow the gap in the next cycle.
But the current index offers little room for optimism on that front. It measures today’s reality, and today’s reality is that closed models are ahead and extending their lead.
For now, the index points one way: proprietary AI is getting stronger, and open alternatives are not keeping pace.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.