Anthropic Commercial APIStatus: active

Claude 3.7 Sonnet

Anthropic's first hybrid reasoning frontier model with dynamic thinking budget control and state-of-the-art coding capabilities.

Released: Feb 24, 2025
Last Verified: Aug 10, 2026
Context Window

200k

Tokens

Max Output

128k

Output limit

Architecture

Hybrid Reasoning Transformer

Model family

Parameters

Not Disclosed

Total / Active

Input Token Cost

$3.00

Per 1M tokens

Output Token Cost

$15.00

Per 1M tokens

Model Overview

Claude 3.7 Sonnet is a state-of-the-art hybrid reasoning model that can switch seamlessly between instantaneous standard inference and extended chain-of-thought reasoning. Developers can specify exact thinking token budgets (up to 128k output tokens) or let the model dynamically decide. It establishes new record scores on SWE-bench Verified (70.3%) and LiveCodeBench, excelling at full-stack software development, architectural refactoring, complex tool execution, and multimodal visual analysis.

Developer Implementation Notes

Supports thinking parameter `thinking: { type: "enabled", budget_tokens: 2048 }`. Prompt caching offers 90% discount on cache hits ($0.30/MTok). Best-in-class performance in agentic workflows when combined with Claude Code CLI and tool calling.

Key Strengths

  • Top-tier coding and software engineering capability (70.3% on SWE-bench Verified)
  • Flexible thinking budget control allowing latency vs reasoning depth tradeoffs
  • Large 128k maximum output token capability
  • 90% prompt caching discount ($0.30 / MTok)

Limitations & Boundaries

  • Proprietary model; cannot run on-premises without cloud agreement
  • Higher latency when extended reasoning budget is maximized

Capabilities & Modalities

Modalities:text, image
Reasoning Chains:Supported
Tool / Function Calling:Supported
Structured Outputs:Supported
Fine-Tuning:Supported
Local Deployment:Cloud Only

Best Production Use Cases

  • Full-stack software engineering and automated bug resolution
  • Complex multi-step autonomous AI agent scaffolding
  • Architectural code refactoring and multi-file debugging
  • Deep mathematical and algorithmic research

Verified Benchmark Results

Standardized evaluations with methodology notes and authoritative citation links.

View all industry benchmarks →
Coding Official
SWE-bench Verified

70.3%

Pass@1, Thinking mode enabled with Claude Code scaffolding and 64k token thinking budget

Coding Official
LiveCodeBench

70.3%

Pass@1, 0-shot with extended thinking budget

Reasoning & Math Official
GPQA Diamond

78.4%

Zero-shot with Chain-of-Thought thinking budget

Reasoning & Math Official
AIME (2024/2025)

80.0%

Pass@1 with 64k thinking budget

General Knowledge & Frontier Official
Humanity's Last Exam (HLE)

22.8%

Zero-shot with thinking budget on CAIS Humanity's Last Exam

Multimodal Official
MMMU

72.5%

Pass@1 multimodal visual reasoning

Pricing & Inference Cost Calculator

Token Cost Estimator – Claude 3.7 Sonnet

Calculate projected inference spend with prompt caching

Quick Workload Presets

Input Tokens / Req2,000
100100k200k
Output Tokens / Req500
5016k32k
Number of Requests1,000
125k50k
Cache Hit Rate0%
0% (No cache)50%90% (Max)
Total Input Spend

$6.0000

2.00M input tokens

Total Output Spend

$7.5000

0.50M output tokens

Estimated Total Cost

$13.50

Avg: $0.01350 / req

Rates: $3.00 in / $15.00 out per million tokens (Anthropic API)Last verified: Aug 10, 2026

Deployment & Integration Options

Anthropic APIcloud api
npm install @anthropic-ai/sdk\n\nconst client = new Anthropic();\nconst response = await client.messages.create({\n  model: "claude-3-7-sonnet-20250219",\n  max_tokens: 4096,\n  thinking: { type: "enabled", budget_tokens: 2048 },\n  messages: [{ role: "user", content: "Write a high-performance LRU cache in Rust." }]\n});
Hardware: Serverless APIProvider Console

Compare with Similar Models

Google DeepMind

Gemini 2.0 Pro (Experimental)

Google's flagship intelligence model engineered for complex coding, mathematical proofs, and 2M token context.

Compare Claude 3.7 Sonnet vs Gemini 2.0 Pro (Experimental)
OpenAI

OpenAI o3-mini

Next-generation cost-efficient reasoning model optimized for STEM, competitive math, and coding.

Compare Claude 3.7 Sonnet vs OpenAI o3-mini
Google DeepMind

Gemini 2.0 Flash

Google's high-speed multimodal workhorse with 1M token context, native tool use, and real-time audio/video streaming.

Compare Claude 3.7 Sonnet vs Gemini 2.0 Flash

Data Accuracy & Verification Notice

AI model specifications, pricing records, and benchmark metrics published on this platform are compiled directly from authoritative sources (official provider documentation, research papers, and verified evaluation harnesses). Benchmark results reflect specific test harnesses and prompting methodologies; scores are not directly comparable across differing evaluation setups.