Model

Claude 3 Opus

CLOSED

Anthropic · Claude 3 family · released Mar 4, 2024

Claude 3 flagship; first Claude to top GPT-4 on major public benchmarks, with vision and 200K context.

ReasoningCodingVisionFunction callingTool useAgentic

1081.0

Elo · rank #193

Parameters

Undisclosed

Active params

—

Context

200K tokens

Architecture

Dense multimodal Transformer (parameter count undisclosed)

License

Proprietary (Anthropic Commercial Terms of Service)

Languages

—

API price (in/out)

$15 / $75

Modalities

text · vision

Benchmark results

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

AI2DVision88.1%#26

best: Molmo 72B · 96.3%

AMCMath42.0%#5

best: Falcon-H1 34B · 69.4%

ARC-ChallengeReasoning96.4%#3

best: Llama 3.1 405B · 96.9%

BIG-Bench HardReasoning86.8%#11

best: ERNIE 4.5 300B-A47B · 94.3%

ChartQAVision80.8%#33

best: MiniMax-VL-01 · 91.7%

DocVQAVision89.3%#35

best: Qwen2-VL-72B · 96.5%

DROPReasoning83.1#19

best: Hunyuan-T1 · 93.1

GPQA DiamondReasoning50.4%#189

best: GPT-5.6 · 94.6%

GSM8KMath95.0%#9

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning95.4%#1

best: this model · 95.4%

HumanEvalCoding84.9%#43

best: Claude Opus 4.5 · 99.4%

LogicKorHuman preference8.88#7

best: GPT-4o · 9.33

MATH-500Math60.1%#111

best: GPT-5 · 99.4%

MathVistaVision50.5%#53

best: Seed 2.1 Pro · 90.7%

MBPPCoding86.4%#3

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

MGSMMath90.7%#10

best: OpenAI o4-mini · 93.7%

MMLUKnowledge86.8%#38

best: OpenAI o3 · 92.9%

MMMLUKnowledge79.1%#28

best: Gemini 3.1 Pro · 92.6%

MMMUVision59.4%#79

best: Claude Fable 5 · 89.3%

PubMedQAKnowledge75.8%#5

best: Palmyra Med 70B · 79.6%

RACE-HReasoning92.9%#1

best: this model · 92.9%

RakudaHuman preference8.75#5

best: PLaMo-100B · 9.73

SWE-bench VerifiedCoding22.0%#113

best: Claude Fable 5 · 95.0%

WinoGrandeReasoning88.5%#5

best: PaLM 2 · 90.9%

Run it locally

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

Input / M tok

$15

Output / M tok

$75

Claude 3 family

Elo progression across releases

Claude 3 OpusMar 20241081.0 Claude 3 SonnetMar 2024821.3 Claude 3 HaikuMar 2024707.0

Resources

DOCSdocs.claude.com

API price $15/$75 · each benchmark row carries its own source badge (see methodology)