Artificial Analysis Launches Progress Index for Language Model Capability Since 2022

Artificial Analysis announced the Artificial Analysis Progress Index, a single capability scale for every language model since 2022, launched alongside Intelligence Index v5. The Progress Index uses item response theory to estimate one general capability score per model across all evaluations, including retired or excluded tasks, enabling comparisons across time. Since reasoning models arrived, the frontier has risen roughly 65 to 70 points a year, about double the 2023 pace, with the highest score among released models reaching 352 for Gemini 4 Carbon in 2026.
Event version 1.
Key insights
Progress Index enables cross-time comparison
The Progress Index uses item response theory to estimate one general capability score for each model from underlying tasks in all evaluations, including retired or excluded ones. The Intelligence Index remains the primary measure of current model intelligence and its scores from different versions cannot be compared.
Frontier growth has roughly doubled since reasoning models
The highest score among released models rose 35 points in 2023, 67 in 2024, 68 in 2025 and 56 so far in 2026. Since reasoning models arrived, the frontier has risen about 65 to 70 points a year, about double the 2023 pace.
Open weights models trail proprietary models by about four months
Since January 2025, each new highest open weights score came 2-8 months after a proprietary model first reached it, with a median of about 4 months. DeepSeek V4.1 Flash 1017 is the highest-scoring model with released weights at 324.
New frontier scores are set more frequently
The highest score among released models was raised on 2 release dates in 2023, 4 in 2024, 9 in 2025 and 11 so far in 2026, about once every four weeks this year.
Leadership at the frontier has become a close contest
OpenAI held the number one position for the first 27 months measured since November 2022. In 2026 there have been rapid changes in the leader between OpenAI, Anthropic and Google.