models.fyi

Side by side, same effort tier

Pick up to four models. Every column is evaluated at the tier you choose (resolved to the nearest tier each model supports), so the comparison is what you'd actually deploy.

MetricGPT-5.6 Luna
Low
Gemini 3.1 Pro
Low
GLM-5.2
Low
Intelligence Index59.063.761.3
Cost / typical task$0.0064$0.016$0.0032
$/M in · out$1 · $6$2 · $12$0.60 · $2.20
Context window1.05M1M200K
Median speed230 tok/s130 tok/s95 tok/s
MMLU-Pro79.082.780.8
GPQA Diamond61.568.363.8
Humanity's Last Exam18.025.519.5
AIME 202551.056.353.4
SWE-bench Verified57.661.661.6
LiveCodeBench63.267.264.8
τ²-bench60.164.763.9
IFEval85.587.485.5

Hover a benchmark name for what it measures — full plain-language guide on /benchmarks· preview data.