AISLE

Leaderboard

Overall and category scores for the AISLE evaluation.

Score range: 0–100Website snapshot: 27 Oct 2026

These are results from a fixed research cohort, not a live ranking of today’s models. The two indices use different task sets and must not be compared directly.

AISLE · 7 models

#ModelOverallCategory contributions

0255075100

Equal category weights. Select a category above or a model for details.

Detailed scores

Order rows by Select a model to view category scores.

Shared-task index

Equal weight across optimization, safety and adaptive robustness.

0–100

Study scores from 0 to 100. Rank refers to the aggregate score.
Rank	Model	Overall / 100	Optimization	Safety	Robustness	Multimodal
01	OpenAI	44.73	25.95	33.33	74.92	33.86
02	Z.ai	32.27	23.80	4.44	68.57	N/A
03	Qwen	31.91	26.13	4.44	65.16	17.71
04	Anthropic	23.91	10.27	8.89	52.58	25.27
05	Google	19.25	28.47	0.00	29.28	20.42
06	DeepSeek	17.05	13.83	0.00	37.32	N/A
07	Moonshot AI	13.91	7.00	0.00	34.73	22.62

Rank uses the overall score. In shared tasks, multimodal is displayed but excluded from the overall score. N/A means not evaluated.

AISLE · 5 models

#ModelOverallCategory contributions

0255075100

Equal category weights. Select a category above or a model for details.

Detailed scores

Order rows by Select a model to view category scores.

Full-task index

Equal weight across all four categories, restricted to five models with full coverage.

0–100

Study scores from 0 to 100. Rank refers to the aggregate score.
Rank	Model	Overall / 100	Optimization	Safety	Robustness	Multimodal
01	OpenAI	42.02	25.95	33.33	74.92	33.86
02	Qwen	28.36	26.13	4.44	65.16	17.71
03	Anthropic	24.25	10.27	8.89	52.58	25.27
04	Google	19.54	28.47	0.00	29.28	20.42
05	Moonshot AI	16.09	7.00	0.00	34.73	22.62

Rank uses the overall score. In shared tasks, multimodal is displayed but excluded from the overall score. N/A means not evaluated.

Scores are equally weighted category indices, not success rates. Shared tasks compare seven models across three categories; full tasks compare five models across four categories. N/A results are excluded, not scored as zero. Scoring details ↗