Claude 3 Sonnet

// Overview

Claude 3 Sonnet is the original Sonnet model, announced on March 4, 2024 as the balanced middle tier of the Claude 3 family, between Claude 3 Opus and Claude 3 Haiku. It introduced the Opus/Sonnet/Haiku naming that every later Claude generation has used.

It was the first Claude model with vision, able to process photos, charts, graphs and technical diagrams alongside text. On standard benchmarks it scored 79.0% on MMLU, 40.4% on GPQA, 73.0% on HumanEval, and 92.3% on GSM8K, positioning it as a strong, affordable model for scaled enterprise deployments.

Claude 3 Sonnet has a 200K token context window, up to 4,096 output tokens, an August 2023 knowledge cutoff, and supports text and image input, at $3 / $15 per million tokens. Its model ID is claude-3-sonnet-20240229; the snapshot was retired on July 21, 2025.

// Benchmarks

1. MMLU79%
2. GPQA40.4%
3. HumanEval73%
4. GSM8K92.3%

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

// Pricing

Pricing source ↗

// Strengths

• Introduced the Opus/Sonnet/Haiku tier naming for the Claude family
• First Claude model with vision (charts, diagrams, photos)
• 200K token context window at an affordable mid-tier price
• Balanced intelligence and speed for enterprise scale
• Solid math and coding scores (92.3% GSM8K, 73.0% HumanEval)

// Best for

• General-purpose text generation and summarization
• Vision tasks over charts, graphs and technical diagrams
• Knowledge-base Q&A and enterprise content workflows
• Cost-sensitive deployments needing a balanced model
• Math and coding assistance at moderate difficulty

// How to access

// Claude Sonnet — every version

The full lineage of the Claude Sonnet line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.