PublicAI Index

@PublicAIData86k followers

The LLM benchmark aggregator — the world's most comprehensive and robust AI index. Built by @PublicAI_

Open on X

Top outlier

10.1×

Best organic post

Peak views

26k

Single post

Bookmarks

Across 9 organic posts

Last scan

Oct 17

Community cache

Posts 9 posts

#	Date	Post	↓ Outlier	Views	Likes	Reposts	Bookmarks
1	Sep 8	1/ We just killed benchmark overfitting in LLM evaluation. Every launch ships with a table the vendor picked. The PublicAI Index is one score from every recognised public leaderboa	10.1×	26k	23	7	3
2	Oct 1	Gemini 4 Argon shipped today and it is on the Index at No. 4, index 66.9. Placed in 17 domains, above par in all 17. The range: - Human preference: No. 1 of 342 - Professional: No.	1.4×	3.8k	2	0	0
3	Oct 14	Muse Wonder is on the Index at No. 7, index 64.1. Placed in 23 domains, above par in 21. The range: - Cybersecurity: No. 1 of 46 - Human preference: No. 2 of 345 - Legal: No. 4 of	1.1×	2.9k	0	0	0
4	Sep 27	The Index now has a Decisions column. It covers two things most leaderboards do not measure: routing and classification, meaning whether a model picks the right path for a task, an	1.1×	2.9k	5	0	1
5	Oct 5	Index Weekly just dropped. Snapshot of how the LLM ranking moved since last week. Three new names in the Top 10: GPT-6.1 Sol @OpenAI: No. 4, index 67.3 Gemini 4 Argon @GoogleDeepMi	1.0×	2.6k	3	0	0
6	Oct 13	OpenAI's flagship and its mid tier are 0.3 points apart on the Index. - GPT-6 Astra: No. 3, index 67.8 -GPT-6.1 Sol: No. 4, index 67.5 - GPT-6 Luna: No. 42	0.9×	2.4k	2	0	0
7	Sep 24	Claude Opus 5.5 is on the Index at No. 2, index 69.3. Opus 5.5 is placed in 19 domains, above par in 18. It leads seven of them, including: - Knowledge work: No. 1 of 179 - General	0.8×	2k	3	0	0
8	Oct 5	Full table:	0.7×	1.9k	2	0	0
9	Sep 23	GPT-6 Sol shipped yesterday. It is on the Index today at No. 11, index 62.2. Aggregated from public leaderboards: @ArtificialAnlys , LiveBench, @vals_ai and Vectara.	0.7×	1.8k	2	0	1
10	Sep 22	Grok 4.7 shipped yesterday. It is on the Index today at No. 19. @xai calls it a notable improvement over Grok 4.6. On the overall index, the two sit a tenth of a point apart: Grok	0.6×	1.7k	3	0	0
11	Oct 1	Note: 5 of 19 sources so far, three of them from one publisher. Expect the number to move.	0.6×	1.7k	1	0	0
12	Oct 13	Note: Aggregated from public arenas and leaderboards.	0.5×	1.4k	0	0	0
13	Oct 1	Learn more:	0.5×	1.2k	0	0	0
14	Oct 14	Note: 7 of 21 sources on Wonder, from four publishers. The Index refreshes twice a day, so the ranking may move. Learn more:	0.4×	1.1k	0	0	0
15	Sep 27	Blog: Index:	0.1×	367	0	0	0
16	Sep 24	Note: 6 of the 18 sources we read have scored it, three of them from one publisher. Expect the number to move.	0.1×	195	0	0	0
17	Sep 27	Note: one publisher feeds this column so far, and every row shows 1/1. We aggregate it z-scored and shrinkage-weighted, the same method as every other column.	0.1×	187	0	0	0
18	Sep 23	Note: 6 of the 18 sources we read have scored it, and three of those six come from one publisher.	0.0×	131	0	0	0
19	Sep 22	Note: Grok 4.6 has 10 boards behind it. Grok 4.7 has 6 so far, so expect the overall number to move.	0.0×	89	1	0	0
20	Sep 24	Note: 6 of the 18 sources we read have scored it, three of them from one publisher. Expect the number to move.	0.0×	57	0	0	0
21	Sep 24	Note: 6 of the 18 sources we read have scored it, three of them from one publisher. Expect the number to move.	0.0×	51	0	0	0

Scan another creator

Run this on any account.

XOutlier builds this table for every public profile you open on x.com, with live data and every post, not just the top 50.

Add to Chrome Download .zip