Model

GPT-4 Turbo

CLOSED

OpenAI · GPT-4 family · released Nov 6, 2023

128K-context, cheaper GPT-4 announced at OpenAI's first DevDay; current default snapshot gpt-4-turbo-2024-04-09 carries a Dec 2023 knowledge cutoff.

ReasoningCodingVisionFunction callingTool useAgentic

1142.7

Elo · rank #183

Parameters

Undisclosed

Active params

Undisclosed

Context

128K tokens

Architecture

Transformer; widely reported (unconfirmed) GPT-4-generation mixture-of-experts

License

Proprietary

Languages

—

API price (in/out)

$10 / $30

Modalities

text · vision

Benchmark results

Bar shows position within the tracked field; marker = field best

Arena-HardHuman preference82.6%#23

best: Qwen3-235B-A22B (Non-Thinking) · 96.1%

Arena EloHuman preference1251#37

best: Claude Fable 5 · 1505

DROPReasoning86#13

best: Hunyuan-T1 · 93.1

GPQA DiamondReasoning49.3%#192

best: GPT-5.6 · 94.6%

HumanEvalCoding88.2%#29

best: Claude Opus 4.5 · 99.4%

LiveCodeBenchCoding28.7%#156

best: DeepSeek-V4-Pro (Think Max) · 93.5%

LogicKorHuman preference9.17#4

best: GPT-4o · 9.33

MATH-500Math73.4%#83

best: GPT-5 · 99.4%

MGSMMath89.6%#14

best: OpenAI o4-mini · 93.7%

MiniF2F (test)Math23.0%#2

best: DeepSeek-Prover-V2 671B · 88.9%

MiniF2F (valid)Math25.4%#4

best: DeepSeek-Prover-V2 671B · 90.6%

MMLUKnowledge86.5%#42

best: OpenAI o3 · 92.9%

MT-BenchHuman preference9.32#3

best: Hunyuan-Large (A52B) · 9.4

RakudaHuman preference9.18#3

best: PLaMo-100B · 9.73

SimpleQAKnowledge24.2%#20

best: GPT-4.5 · 62.5%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

$10

Output / M tok

$30

GPT-4 family

Elo progression across releases

GPT-4Mar 20231037.4 GPT-4 TurboNov 20231142.7

Resources

DOCSdevelopers.openai.com/api/docs/models/gpt-4-turbo APIAvailable via OpenAI API

API price $10/$30 · each benchmark row carries its own source badge (see methodology)