← Models

Claude Sonnet 5

Anthropic · Claude 5

Anthropic's balanced Sonnet model for agentic coding, tool use, and production assistants.

Part of Claude Sonnet family · Other versions: Claude Sonnet 4.6 , Claude Sonnet 4.5
Type
multimodal
Context
1M tokens
Max Output
128K tokens
Status
current
Input
$2/1M tok
Output
$10/1M tok
API Access
Yes
License
proprietary
reasoning coding agentic vision tool-use long-context balanced
Released June 2026 · Updated July 6, 2026

Overview

Freshness note: Model capabilities, limits, pricing, and availability can change quickly. This profile is a point-in-time snapshot last verified on July 6, 2026.

Claude Sonnet 5 is Anthropic’s June 30, 2026 successor to Claude Sonnet 4.6. It is now the Sonnet-family page to cite when a workflow needs a balanced Claude model rather than the maximum-cost premium route. Anthropic positions it as the most agentic Sonnet model so far, with stronger planning, tool use, coding, browser and terminal workflows, and follow-through than earlier Sonnet releases.

The important routing change is simple. Sonnet 4.6 remains useful for pinned deployments, but Sonnet 5 is the current Sonnet default across Claude products and the Claude Platform.

Capabilities

Sonnet 5 is built for the kind of multi-step work that used to push teams toward Opus sooner: repository changes, tool-using agents, long-context document analysis, computer-use style tasks, and knowledge-work workflows where the model needs to keep a plan alive. Anthropic says it narrows the performance gap with Opus 4.8 while keeping a lower price point.

It also changes the practical economics of Claude routing. At launch, the introductory API price is lower than the standard Sonnet 4.6 price, while the model offers a larger 128K max output window and a 1M context window in the current model overview.

Technical Details

Current official anchors:

  • API ID: claude-sonnet-5
  • 1M-token context window
  • 128K max output
  • text and image input, text output
  • adaptive thinking support
  • availability across Claude plans, Claude Code, and Claude Platform

Anthropic notes that Sonnet 5 uses a newer tokenizer. The same input can map to more tokens than with Sonnet 4.6, so teams should measure real workloads instead of comparing only headline per-token prices.

Pricing & Access

Introductory API pricing through August 31, 2026:

  • Input: $2 per 1M tokens
  • Output: $10 per 1M tokens

Starting September 1, 2026, standard pricing becomes:

  • Input: $3 per 1M tokens
  • Output: $15 per 1M tokens

Prompt caching, cache hits, batch API discounts, cloud-provider pricing, and data-residency multipliers can change the real bill. Anthropic’s pricing docs also call out tokenizer-related effective-cost differences, so production migrations should run cost tests before switching large workloads.

Best Use Cases

Use Sonnet 5 as the default Claude route for most serious production work: coding agents, code review, document-heavy analysis, customer-support assistants, operations workflows, and team tools that need strong reasoning without Opus or Fable pricing.

Move up to Opus 4.8, Fable 5, or Mythos 5 only when the task needs a higher ceiling and the access, risk, and cost profile are justified.

Comparisons

  • Claude Sonnet 4.6 (Anthropic): Previous Sonnet default, retained for pinned deployments and behavior regression tests.
  • Claude Opus 4.8 (Anthropic): Higher-ceiling premium route for harder agentic coding and complex reasoning.
  • Claude Fable 5 (Anthropic): More capable and more expensive broadly released model with stronger safeguards.
  • GPT-5.5 (OpenAI): Comparable OpenAI-side flagship for teams centered on ChatGPT, Codex, and Responses API workflows.