Model

Claude 1

CLOSED

Anthropic · Claude 1 family · released Mar 14, 2023

Anthropic's first public model; brought Constitutional AI training to a commercial product. Retired Nov 2024.

ReasoningCodingVisionFunction callingTool useAgentic

758.5

Elo · rank #256

Parameters

Undisclosed

Active params

—

Context

9K tokens

Architecture

Dense transformer (exact architecture undisclosed)

License

Proprietary

Languages

—

API price (in/out)

$11.02 / $32.68

Modalities

text

Benchmark results

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

ARC-ChallengeReasoning90.0%#26

best: Llama 3.1 405B · 96.9%

Arena EloHuman preference1195#42

best: Claude Fable 5 · 1505

GSM8KMath85.2%#67

best: Llama 3.1 405B · 96.8%

APIBench (HuggingFace)Agents6.2%#4

best: Gorilla 7B · 71.7%

MMLUKnowledge77.0%#116

best: OpenAI o3 · 92.9%

OpenBookQAReasoning90.8%#1

best: this model · 90.8%

RACE-HReasoning88.8%#2

best: Claude 3 Opus · 92.9%

APIBench (TensorHub)Agents9.2%#4

best: Gorilla 7B · 83.8%

APIBench (TorchHub)Agents18.8%#4

best: Gorilla 7B · 59.1%

TriviaQAKnowledge86.7%#7

best: Sarvam-1 (2B) · 90.6%

Run it locally

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

Input / M tok

$11.02

Output / M tok

$32.68

Claude 1 family

Elo progression across releases

Claude 1Mar 2023758.5

Resources

DOCSwww.anthropic.com/news/introducing-claude

API price $11.02/$32.68 · each benchmark row carries its own source badge (see methodology)