GPT-4

OpenAI · openai/gpt-4

ProviderOpenAI
Familygpt
Typellm-chat
Statusactive
Released2023-11-06
Updated2024-04-09

Reported benchmark scores

These scores carry one collection date for the whole card (2026-04) and a source list rather than a source per score, so they cannot be attributed individually. They are shown as reported and are not verified evidence. Benchmark pages carry the reviewed, dated evidence where it exists.
BenchmarkCatalogue standingScore
AI2D (AI2 Diagrams)unassessed95.2
Aider Polyglot Benchmarkunassessed52.4
AlpacaEvalunassessed52.1
Arena Elo — Codingunassessed1240.0
Arena Elo — Mathunassessed1220.0
Arena Elo — Overall (Text)unassessed1230.0
BBQ (Bias Benchmark for QA)unassessed86.5
ChartQAunassessed88.5
DocVQAunassessed94.1
FLORES English-to-Germanunassessed67.3
FLORES English-to-Spanishunassessed71.2
FLORES English-to-Japaneseunassessed61.4
FLORES English-to-Chineseunassessed59.7
GPQA Diamondunassessed39.7
GSM8Kunassessed92.0
HELM Safetyunassessed90.5
HumanEvalunassessed67.0
IFEvalunassessed76.5
LiveCodeBenchunassessed48.3
MATH-500unassessed42.5
MathVistaunassessed68.2
MGSM (Multilingual Grade School Math)unassessed74.5
MMLU-Prounassessed56.2
MMMUunassessed72.5
MT-Benchunassessed9.3
MultiPL-E: C#unassessed76.5
MultiPL-E: Juliaunassessed56.2
MultiPL-E: Kotlin (unconfirmed)unassessed68.2
MultiPL-E: Luaunassessed50.8
MultiPL-E: Perlunassessed46.5
MultiPL-E: PHPunassessed74.2
MultiPL-E: Runassessed52.8
MultiPL-E: Rubyunassessed62.5
MultiPL-E: Scalaunassessed56.8
MultiPL-E: Swiftunassessed61.5
SWE-bench Verifiedunassessed54.6
ToxiGenunassessed93.5
WildBenchunassessed80.5

Data

This card as JSON · Edit on GitHub