Model

Claude 2

CLOSED

Anthropic · Claude 2 family · released Jul 11, 2023

First broadly available Claude with a 100K-token context window and markedly stronger coding/math than Claude 1.3.

ReasoningCodingVisionFunction callingTool useAgentic

821.5

Elo · rank #250

Parameters

Undisclosed

Active params

—

Context

100K tokens

Architecture

Dense decoder-only Transformer (parameter count undisclosed)

License

Proprietary (Anthropic Commercial Terms of Service)

Languages

—

API price (in/out)

$8 / $24

Modalities

text

Benchmark results

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

ARC-ChallengeReasoning91.0%#22

best: Llama 3.1 405B · 96.9%

GSM8KMath88.0%#52

best: Llama 3.1 405B · 96.8%

HumanEvalCoding71.2%#83

best: Claude Opus 4.5 · 99.4%

MMLUKnowledge78.5%#102

best: OpenAI o3 · 92.9%

RACE-HReasoning88.3%#4

best: Claude 3 Opus · 92.9%

TriviaQAKnowledge87.5%#5

best: Sarvam-1 (2B) · 90.6%

TruthfulQAKnowledge69.0%#6

best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%

Run it locally

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

Input / M tok

Output / M tok

$24

Claude 2 family

Elo progression across releases

Claude 2Jul 2023821.5 Claude 2.1Nov 2023565.4

Resources

DOCSdocs.claude.com

API price $8/$24 · each benchmark row carries its own source badge (see methodology)