LANGUAGE MODEL Anthropic Last updated:

Claude 3.7 Sonnet

Hybrid Reasoning Model

The first Claude model that can switch between fast standard responses and slow "extended thinking" mode in a single product — rather than offering two separate models like OpenAI's GPT-4o vs o-series split. Achieved state-of-the-art on real software engineering benchmarks (SWE-bench Verified) and powered Anthropic's first agentic IDE feature, Claude Code, released the same week.

Intelligence
Medium
Cost
High
$3.75 in / $15.00 out
Context
200K
Up to 200,000 tokens
How are Intelligence, Speed & Cost bucketed?
Intelligence and Speed buckets are percentile ranks on Artificial Analysis. Cost buckets are fixed dollar thresholds keyed off output-token price ($/M out).
Intelligence
  • Top 1%≤ 1%
  • Top 5%≤ 5%
  • Top 10%≤ 10%
  • Good≤ 25%
  • Medium≤ 50%
  • Below avg> 50%
Speed
  • Top 1%≥ 345 tok/s
  • Top 5%≥ 237 tok/s
  • Top 10%≥ 196 tok/s
  • Good≥ 146 tok/s
  • Medium≥ 90 tok/s
  • Slow< 90 tok/s
Cost
  • Freeopen weights · self-host
  • Low< $1 / M out
  • Moderate$1–5 / M out
  • High≥ $5 / M out

Why it matters

Claude 3.7 Sonnet was the first model where "AI writes meaningful portions of production codebases autonomously" stopped being a demo and started being a measurable share of real engineering work. The SWE-bench Verified leaderboard from Q1 2025 onward reads like an industry transition.

Core Capabilities

Long Documents
Handles entire codebases, books, and multi-doc RAG.
Multimodal
Combines text, vision, and audio in one model.
Generative
Produces images, video, audio, or other media.
Agent Workflows
Built for tool use and autonomous tasks.

Context Window

200k tokens
≈ 154+ pages
4k Chat 聊天
32k Long docs 长文档
128k Books 整本书
400k Multi-doc 多文档
1M Codebase 整个代码库
10M
200k

Availability

API
Available
Product / App
Available
Open Source
Not released
Enterprise
Contact sales

Pricing Model

Pay per token
Input and output billed separately.
Pay-per-token

Capability / Performance

Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).

Lower 20% Upper 80% This model
Reasoning
AA Intelligence Index · scaled to 10
1.7
5.6
5.0
Coding
SciCode · scaled to 10
1.8
4.3
4.0
Agentic tasks
Terminal-Bench Hard · scaled to 10
0.2
3.6
2.1
Context / memory
Context window size · log-scaled
6.0
9.0
6.7
Cost efficiency
Input price ($/M tokens) · cheaper scores higher
6.2
10.0
4.5
Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.

What it feels like

  • First hybrid-reasoning model — same checkpoint switches between fast standard mode and extended thinking
  • 70.3% on SWE-bench Verified with custom scaffold — industry-leading at release; jumped from 62.3% baseline
  • Output token limit jumped 8x to 64K (128K beta) — made long-form code/docs viable in one response
  • Anthropic showed the chain-of-thought to users — competitor o1 hides it
  • Same price as 3.5 Sonnet ($3 in / $15 out, includes thinking tokens) — extended thinking is essentially free
  • Bug-find-and-fix loops finally feel reliable: read report → navigate codebase → root cause → working patch

Best use cases

  • Agentic coding with Claude Code CLI — the model designed for this surface
  • Math proofs, multi-step logic, and scientific analysis where extended thinking pays off
  • Long-form code generation, technical writing, or analysis (up to 128K output)
  • Teams that want one model toggling between fast and slow modes via prompt

Tools to try

Not ideal for

  • Casual chat or copy that doesn't need reasoning — extended thinking adds latency without benefit
  • Hard reasoning leaderboards by late 2025 — Claude 4 / Opus 4.5 have moved well past it
  • Workloads needing 1M context (3.7 Sonnet ships 200K)

Model Evolution

View full evolution tree →