Question
On 2028-06-30, how many organizations will have an independently pretrained model scoring within five points of the top score on the Artificial Analysis Intelligence Index?
We project a median of 3 organizations by mid-2028. Currently, on v5.0.1 (read 2026-11-13), two organizations are inside the five-point band: Anthropic (64.6 with Claude Fable 5.2) and Google (60.0 with Gemini 4 Carbon) Artificial Analysis Intelligence Index Leaderboard LLM Leaderboard - Comparison of AI models from OpenAI, ... Claude Fable 5.2 and Mythos 5.2 cement Anthropic's AI lead .... OpenAI is close behind at 56.9. Historically across 2026, this count has fluctuated between 1 and 5, typically hovering around 3 to 4 as Anthropic, Google, and OpenAI leapfrog one another. Carbon, for instance, held the top spot for only 12 days before Fable 5.2's release AI Insights & Trends. We expect these three to remain the core contenders, continually trading the lead and dictating the competitive threshold. Set against related questions, the distribution has remained largely stable, reinforcing our expectation that the top three labs will likely dominate the frontier with limited crowding.
A critical driver of this count is how Artificial Analysis manages its index. Benchmark saturation mechanically compresses score spreads over time, which would naturally increase the number of qualifying organizations within a fixed 5-point band. However, the maintainers actively counter this by retiring saturated evaluations and adding harder, private test sets. The launch of version 5 in October 2026, which moved 60% of the index to private or held-out sets, abruptly widened the open-weights gap from 3 to 12 points Announcing Artificial Analysis Intelligence Index v5 Artificial Analysis Intelligence Index v5 (major update) Announcing Artificial Analysis Intelligence Index v5. By June 2028, the reading will almost certainly be on version 6 or later, and this active rebasing will likely prevent the field from clustering too tightly at the top.
Moving the count to 4 or 5 requires well-funded challengers like Meta, xAI, or Chinese developers to close the gap. Currently, they are far behind on v5: Meta is 14.5 points back, Moonshot is 18.1 points back, and xAI fell roughly 29 points behind the frontier after the version update https://the-decoder.com/new-artificial-analysis-index-widens-the-gap-between-open-and-closed-models/ Grok 5 vs GPT-6 and Gemini 4: Can xAI Catch Up?. While capital and compute are scaling massively—such as xAI’s targeted 10–15 GW build-out by end-2027 Grok 5 vs GPT-6 and Gemini 4: Can xAI Catch Up? and Alibaba's heavy infrastructure investments June Quarter 2026 Results Conference Call Transcript—translating raw compute into sustained frontier capability is difficult. Furthermore, the strict requirement that qualifying models be independently pretrained excludes developers relying on distills or fine-tunes, keeping the realistic pool of entrants small.
Finally, the specific mechanism proposed by Dwarkesh Patel—that models will diverge significantly by learning from different live experiences—is not operating at the frontier as of late 2026, as no leading systems update their weights in production. If continual learning is deployed by 2028, it is more likely to help capital-rich incumbents pull further away than to enable challengers to catch up. A count of 1 or 2 is therefore a realistic downside scenario if index rebasing is particularly aggressive or if algorithmic breakthroughs allow a single lab to establish a dominant lead. Conversely, counts of 6 or more require a prolonged period of benchmark saturation aligning perfectly with successful capability jumps by multiple tier-two labs, making it a distinct but less likely tail event. Taken together, these considerations put our 10th and 25th percentiles at 1.5 and 2.1 organizations, our median at 3.0, and our 75th and 90th percentiles at 4.1 and 5.4 organizations.
Set against related questions, the distribution remained largely stable, confirming our expectation that the top three labs will likely dominate the frontier with limited crowding.
Ask a followup
Sign in to run · $20 free credit, no card · every claim cited