Model Comparison

Deepseek V3.1 vs GPT 6.2 Astra

Deepseek V3.1 ranks #145 on the AI IQ leaderboard with an estimated IQ of 98, while GPT 6.2 Astra ranks #3 on the AI IQ leaderboard with an estimated IQ of 142. The table below puts the two models side by side across the seven capability dimensions, emotional reasoning, cost, speed, and context window.

Where They Sit on the IQ Bell Curve
Both models highlighted on the standard normal IQ distribution of every tracked model
IQ Dimensions Head to Head
The seven scored Composite IQ dimensions, side by side
Domains Head to Head
Applied domain composite scores where at least one model is measured
Deepseek V3.1DeepSeekGPT 6.2 AstraOpenAI
Rank#145#3
IQ98142
Abstract Reasoning73134
Mathematical Reasoning100153
Academic Reasoning113134
Programmatic Reasoning104150
Computer Use95151
Reliability102137
Knowledge Work96132
Emotional Reasoning——
Effective Cost / 1M Tokens$2.49$28.15
Sticker Price / 1M I/O Tokens$2.26$48.00
Speed121 tok/s—
Context Window—1.1M
Released2025-08-212026-11-17
Open WeightsYesNo