Atlas

Models

← All models

ChatGPT-4o Latest (source-unspecified snapshot)

OpenAI · 2024-08-15

ChatGPT-4o Latest (source-unspecified snapshot) is a benchmark-entry representation of OpenAI's mutable `chatgpt-4o-latest` alias, which pointed to the GPT-4o snapshot used in ChatGPT and changed over time. OpenAI has since removed the alias from the API, and the benchmark sources do not identify which underlying dated snapshot served each result.

Benchmark scores

BenchmarkScore
AA-LCR53.0
AA-LCR v1.156.0
AidanBench890
Artificial Analysis Progress Index v1172
BuzzBench59.5
Fiction.LiveBench - 0 tokens (relevant passages only)87.5
Fiction.LiveBench - 120K tokens65.6
Fiction.LiveBench - 16K tokens66.7
Fiction.LiveBench - 1K tokens66.7
Fiction.LiveBench - 2K tokens63.9
Fiction.LiveBench - 32K tokens63.9
Fiction.LiveBench - 400 tokens83.3
Fiction.LiveBench - 4K tokens63.9
Fiction.LiveBench - 60K tokens55.6
Fiction.LiveBench - 8K tokens66.7
GPQA Diamond (Artificial Analysis avg@5)51.1
HardGeoBench <=100mi43.8
HardGeoBench <=10mi31.3
HardGeoBench Avg Distance1578
HardGeoBench Country Accuracy52.5
Humanity's Last Exam (Text-Only)2.7
LisanBench Difficulty-Weighted94.2
LisanBench Path Length303
MATH-500 (Artificial Analysis source)79.7
MMLU Pro77.3
Ruby LLM Benchmarks — Program Fixer68.7
Ruby LLM Program Fixer — Calendar78.0
Ruby LLM Program Fixer — Parking Garage44.4
Ruby LLM Program Fixer — School Library75.3
Ruby LLM Program Fixer — Vending Machine77.1
Loading Atlas data…