CompareBestAI
Join UsJoin Us
Sonnet 4.6

Sonnet 4.6

by Anthropic PBC

Mid‑tier Claude model with 1M context and $3/$15 API.

Reviewed August 2026by CBAI Editorial Team
8.0/10

Score

4.0 out of 5 · our score
Compare
Developer

Our verdict

A strong, widely available mid‑tier model with 1M context and predictable $3/$15 MTok pricing; still a safe pick in 2026, though new builds should evaluate Sonnet 5.

Overview

Claude Sonnet 4.6 is Anthropic’s mid‑tier Claude model, released February 17, 2026 and currently available as a legacy model. It offers a 1M‑token context window at standard pricing, improved coding and computer‑use performance over prior Sonnet versions, and broad availability across Anthropic’s API and major cloud partners. The model accepts text and images and returns text, supports adaptive thinking, and benefits from cost controls like prompt caching and a 50% Batch API…

Score breakdown

Overall score

8.0/10
Output quality8.0/10
Ease of use8.0/10
Value for money8.0/10
Features & tools8.0/10
API & integrations8.0/10
Support & docs7.5/10

Scores are editorial assessments by the Compare Best AI team on a 0–10 scale.

Expert review

C

CBAI Editorial Team

Compare Best AI · Editorial Team

Claude Sonnet 4.6 strikes a compelling balance of capability, latency, and price. In February 2026 it replaced Sonnet 4.5 as the default mid‑tier Claude, delivering better coding, computer use, and long‑context reasoning. Its 1M‑token context is available at standard rates, and prompt caching plus the Batch API help keep spend predictable. Tooling such as web search/fetch with automatic code execution, memory, and programmatic tool calling are supported on the platform and work well with Sonnet 4.6. As of August 24, 2026, Sonnet 4.6 is marked legacy but remains supported across Anthropic’s API and partner clouds (Amazon Bedrock, Google Cloud’s Vertex AI, Microsoft Foundry). For net‑new projects, Anthropic recommends Sonnet 5, which offers improved performance and lower token pricing; however, Sonnet 4.6 remains a reliable choice for teams with established prompts, evals, and governance tied to the 4.6 generation.

How we tested

0

Tasks evaluated

  • ·Verified model specs (context, outputs, cutoffs, lifecycle) from official model page.
  • ·Verified API and feature pricing (token rates, caching, batch discount, US‑only multiplier) from developer pricing docs.
  • ·Verified availability on partner platforms (Vertex AI; listed platforms on model page).
  • ·Confirmed launch details and default model status for Free/Pro from Anthropic news post.
  • ·Checked compliance posture (SOC 2 Type I/II) and default data‑use policies from Help/Privacy docs.

Method

Document review of Anthropic’s official docs, pricing pages, and launch materials; no hands‑on benchmarking.

Reviewer

CBAI Editorial Team

Plans & pricing

Claude API (usage‑based)

Usage-based (per MTok)/usage

Developers using the Sonnet 4.6 model via API or partner clouds.

  • Sonnet 4.6: $3/MTok input; $15/MTok output.
  • Prompt caching: write 5m $3.75/MTok; write 1h $6/MTok; read $0.30/MTok.
  • Batch API: 50% discount on input and output tokens.
  • Context window: 1M tokens; max output 128K (300K via Batch API).
  • Platforms: Anthropic API, Amazon Bedrock (InvokeModel), Google Cloud (Vertex AI), Microsoft Foundry.
  • US‑only inference available at 1.1x multiplier (see Data residency).

Pro

$20/month

Individuals using claude.ai with higher usage and features.

  • Annual billing: $17/mo ($200/yr).
  • Includes Claude Code, Claude Cowork, Claude Design, Claude Science.
  • More usage than Free; access to more Claude models; projects; Claude for Microsoft 365.
  • Sonnet 4.6 was the default model at launch across Free/Pro.

Max 5x

$100/month

Power users who need 5x Pro usage and higher output limits.

  • Includes everything in Pro, with 5x Pro usage per session.
  • Early access to advanced features; priority access at high traffic times.

Team (Standard seat)

$25/month

Teams of 2–150 needing collaboration and admin controls.

  • Annual billing: $20/seat/mo ($240/yr).
  • All Claude features with more usage than Pro.
  • Monthly or annual billing options.

Team (Premium seat)

$125/month

Teams needing 5x Standard usage per seat.

  • Annual billing: $100/seat/mo ($1,200/yr).
  • Everything in Standard with 5x usage.

Pricing may vary by region. Always verify on the vendor's website.

Feature comparison

  • 1M‑token context window at standard pricing; max output 128K tokens (300K via Batch API).
  • Usage‑based API rates: $3/MTok input, $15/MTok output for Sonnet 4.6.
  • Prompt caching supported ($3.75/MTok write 5m, $6/MTok write 1h, $0.30/MTok read).
  • Batch API with 50% discount on input and output tokens.
  • Adaptive thinking (extended thinking deprecated for this model).
  • Multimodal input: text and images → text output.
  • Knowledge cutoff: reliable Aug 2025; training data cutoff Jan 2026.
  • Available on Anthropic API, Amazon Bedrock (InvokeModel), Google Cloud (Vertex AI), and Microsoft Foundry.
  • Web search and fetch tools available on API; code execution, memory, programmatic tool calling generally available.
  • Data residency option via inference_geo (US‑only multiplier) on supported models.

Is it right for you?

Good fit for

Large‑context coding and codebase understanding

1M‑token context lets Sonnet 4.6 ingest repositories or multi‑file diffs for accurate edits and refactors.

Agentic workflows and computer use

Supports tool use and computer‑use improvements noted at launch for RPA‑style tasks.

Batch processing at scale

Batch API halves token costs for asynchronous, high‑volume jobs.

Less suited for

Deepest reasoning tasks

Anthropic recommends Opus for tasks demanding the strongest reasoning accuracy.

Greenfield builds needing latest model economics

New projects may prefer Sonnet 5 ($2/$10 MTok) per Anthropic’s current lineup.

User reviews

4.0

Editorial score

5
48%
4
28%
3
6%
2
4%
1
0%

Distribution is estimated from our editorial score. Verified user reviews coming soon.

Use cases

Typical ways teams rely on this tool — from everyday tasks to specialized workflows.

  • Code assistance and refactoring across large codebases (1M context).
  • Long‑document Q&A, retrieval, and analysis over PDFs, tables, and research corpora.
  • Agentic workflows and computer‑use/RPA tasks with tool use enabled.
  • High‑volume offline jobs via Batch API with 50% input/output discounts.

Integrations

Reported connectors

Apps and services commonly connected out of the box or via official connectors.

  • Amazon Bedrock
  • Google Cloud Vertex AI
  • Microsoft Foundry
  • Claude Platform on AWS (AWS Marketplace billing)

Privacy & compliance

Security & data practices

Highlights we track from the vendor's documentation. Always confirm current terms, subprocessors, and regional policies on their official site.

  • Anthropic states SOC 2 Type I & II compliance
  • by default, commercial content from Claude Platform/API isn’t used to train models (opt‑in only). Zero Data Retention is available on eligible endpoints/models. US‑only inference supported via pricing multiplier. See Trust/Help and Pricing docs.

Details

Category

Developer

Price

  • API usage: $3/MTok input, $15/MTok output
  • Pro $20/mo ($17/mo billed annually).

Free version

Yes

Best for

Teams building code assistants, long‑document analysis, and agentic RPA/computer‑use workflows who want broad platform availability and predictable usage pricing.

Frequently asked questions

Sonnet 4.6

Sonnet 4.6

Developer

Ready to get started?

Visit the Sonnet 4.6 website to explore plans and start your free trial.

Was this page helpful?

Compare Best AI may earn a commission when you click links on this page. This does not influence our editorial scores or recommendations. Advertiser disclosure