Compared to 2025, the indicators have increased significantly, which reflects the rapid development of the capabilities of modern models.
In April 2026, within the TrackingAI benchmark, performed for Mensa Norway, two models at once - Grok-4.20 Expert Mode and OpenAI GPT 5.4 Pro (Vision) - showed the same highest result, scoring 145 points each and taking first place.
It is noted that in the premium segment of AI solutions, competition has noticeably intensified: now the gap between the leaders is only a few points. Compared to 2025, the indicators have increased significantly, which reflects the rapid development of the capabilities of modern models for logical analysis and recognition of visual patterns.
Of particular note is the density of results in the upper part of the rating. The leaders share the first place with the same score, while Gemini 3.1 Pro Preview is next, scoring 141 points. The minimal difference between the positions emphasizes that even minor changes in the results can affect the balance of power.
Noticeable and overall progress compared to last year: if in 2025 the maximum score was 135 points, then in 2026 it reached 145. This indicates the high speed of development of artificial intelligence technologies and their ability to quickly improve results in such tests.
However, not all participants demonstrate the same dynamics. In the considered dataset, the best model of the company Mistral was at the bottom of the list, receiving 97 points, which is significantly lower than the leaders.
It is reported that the TrackingAI benchmark is based on the open test Mensa Norway, which contains 35 tasks for recognizing visual patterns. Models without machine vision capabilities are asked questions aloud, while systems with visual analysis support receive images directly.
Thus, the results should be considered primarily as a comparative study, and not a final assessment of the intelligence level of the models. Since the test is focused on visual perception, the final indicators may depend on the format of the tasks.
TOP-10 smartest AI models in points
Grok-4.20 Expert Mode 145
OpenAI GPT 5.4 Pro (Vision) 145
Gemini 3.1 Pro Preview 141
OpenAI GPT 5.4 Thinking (Vision) 139
OpenAI GPT 5.3 136
Grok-4.20 Expert Mode (Vision) 133
OpenAI GPT 5.4 Thinking 133
Meta Muse Spark 133
Gemini 3.1 Pro Preview (Vision) 132
Qwen 3.5 130
Source: https://dengi.ua/ua/business/9760112-najrozumnishi-nazvano-top-10-modelej-shi-u-2026-rotsi
