← Models

Claude Sonnet 4.6

Anthropic · Claude 4

Anthropic's previous balanced Sonnet model, retained for pinned deployments and behavior comparison.

Part of Claude Sonnet family · Other versions: Claude Sonnet 5 , Claude Sonnet 4.5
Type
multimodal
Context
1M tokens
Max Output
64K tokens
Status
legacy
Input
$3/1M tok
Output
$15/1M tok
API Access
Yes
License
proprietary
reasoning coding balanced vision tool-use
Released February 2026 · Updated July 6, 2026

Overview

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

Claude Sonnet 4.6 launched on February 17, 2026 and was Anthropic’s balanced Sonnet-tier model until Claude Sonnet 5 arrived on June 30, 2026. It remains relevant for pinned deployments, regression testing, and teams that need stable behavior while they evaluate Sonnet 5’s pricing, tokenizer, and agentic changes.

The practical reason this page still matters is continuity. Sonnet 4.6 was not just a routine patch over 4.5; it improved coding, design, long-context reasoning, and computer use while keeping Sonnet-tier pricing. For new Claude builds, however, Sonnet 5 is now the better starting point.

Capabilities

Anthropic’s current launch and product pages emphasize a wide capability spread:

  • Strong coding and repo-scale implementation work, with Anthropic explicitly calling out improved consistency and instruction following.
  • Better computer use behavior for browser and desktop-style workflows where the model needs to navigate software without custom APIs.
  • Long-context reasoning for document-heavy or codebase-heavy work.
  • Agent planning and tool-enabled workflows for production assistants and automations.
  • Solid multimodal understanding for images and documents alongside text reasoning.

In practical routing terms, Sonnet 4.6 is now the previous-generation balanced Claude tier. It still covers substantial coding and assistant work, but Sonnet 5 is the current default and has a larger max output window in the latest model overview.

Technical Details

Anthropic’s current published specs and docs support these reference points:

  • 1M-token context window in the current model overview.
  • Anthropic’s Sonnet product page still notes that the 1M context window is currently available in beta on the API only.
  • 64K max output tokens in the synchronous Messages API.
  • Up to 300K output tokens in the Message Batches API with Anthropic’s documented output-300k-2026-03-24 beta header.
  • Text and image input with text output, plus tool-use support.
  • Extended thinking and adaptive thinking are both supported.

The 1M-token profile matters operationally because it gives Sonnet a credible long-context lane for larger codebases and document sets. Teams should still verify whether their exact route is Claude API, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry, or Claude product UI before assuming identical availability.

Pricing & Access

Anthropic states that base pricing remains the same as Sonnet 4.5:

  • Input: $3 per 1M tokens
  • Output: $15 per 1M tokens
  • 5-minute cache writes: $3.75 per 1M tokens
  • 1-hour cache writes: $6 per 1M tokens
  • Cache hits and refreshes: $0.30 per 1M tokens
  • Batch API input: $1.50 per 1M tokens
  • Batch API output: $7.50 per 1M tokens

Current Anthropic pricing docs list 1M context at standard Sonnet 4.6 pricing. US-only inference through Anthropic’s first-party data-residency option applies a 1.1x multiplier.

Access options:

  • Claude.ai on web, iOS, and Android
  • Anthropic API (direct)
  • Amazon Bedrock
  • Google Cloud Vertex AI
  • Microsoft Foundry
  • Claude consumer and workspace products, including Claude Code-connected workflows and broader Claude desktop surfaces where plan access allows

Best Use Cases

Use Sonnet 4.6 for compatibility-sensitive Claude deployments that already rely on its behavior:

  • Existing production assistants pinned to claude-sonnet-4-6
  • Regression testing against earlier Sonnet behavior
  • Long-context workflows that have not yet been re-costed on Sonnet 5
  • Cloud-provider routes where a team has not confirmed Sonnet 5 availability

For fresh builds, start with Claude Sonnet 5. Move up to Opus, Fable, or Mythos only when the task justifies the access and cost profile, or down to Haiku when throughput dominates the decision.

Comparisons

  • Claude Sonnet 5 (Anthropic): Current Sonnet default with stronger agentic positioning, 128K max output, and introductory pricing through August 31, 2026.
  • GPT-5.5 (OpenAI): Stronger single-model option when a team wants one OpenAI route across coding, reasoning, and agentic tasks, but Sonnet remains a compelling Claude-native default.
  • Gemini 2.5 Pro (Google): Strong long-context and multimodal alternative, especially in Google-native stacks, while Sonnet often wins on Anthropic ecosystem fit and coding-centric workflow familiarity.
  • Claude Opus 4.8 (Anthropic): Higher-ceiling premium escalation path for the hardest work, while Sonnet 4.6 still covers a larger share of real production use without Opus-level cost.