The Model Index
Up to four models

GPT-4 Turbo (Apr 2024) vs GPT-3.5 Turbo (Jan 2024)

On the index (GPT-4 = 100), GPT-4 Turbo (Apr 2024) scores 108 and GPT-3.5 Turbo (Jan 2024) 90. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.

Next to GPT-4 Turbo (Apr 2024)
Presets
GPT-4 Turbo (Apr 2024)
10892–125
WeightsClosed
ReleasedApr 2024
Price in / out–
Context–
Core tests taken3 of 9
GPT-3.5 Turbo (Jan 2024)
9072–109
WeightsClosed
ReleasedJan 2024
Price in / out–
Context–
Core tests taken2 of 9

GPT-4 Turbo (Apr 2024) leads GPT-3.5 Turbo (Jan 2024) by 18 points, ahead on 2 of 2 shared core tests.

Test by testthird-party results

GPT-4 Turbo (Apr 2024)GPT-3.5 Turbo (Jan 2024)
0%25%50%75%100%REASONING & KNOWLEDGEGPQA Diamond (Epoch AI)GPQA Diamond (Epoch AI)MMLU (HELM runs)MMLU (HELM runs)MATHSFrontierMath (tiers 1-3)FrontierMath (tiers 1-3)AIME-style maths (OTIS mock)AIME-style maths (OTIS mock)MATH (level 5)MATH (level 5)CODINGWeirdMLWeirdML
Every result as a table
TestGPT-4 Turbo (Apr 2024)GPT-3.5 Turbo (Jan 2024)
Reasoning & knowledge
GPQA Diamond (Epoch AI)46.6%27.2%
MMLU (HELM runs)81.3%67.3%
Maths
FrontierMath (tiers 1-3)0.7%0.0%
AIME-style maths (OTIS mock)6.7%2.2%
MATH (level 5)46.7%11.6%
Coding
WeirdML18.0%3.5%
Task length · METR22 sec

Tests at least two of these models have taken. Bold rows are the core tests.

This page as a chartDownload PNG