← Back to the A-Z glossary
Model · Anthropic · 2026

Claude Sonnet 4.6

Claude Sonnet 4.6 is Anthropic's balanced model in 2026: the one in the middle, between Haiku (small, fast) and Opus (big, slow, pricey). Smart enough to write, analyze a document or code, fast enough that you're not left waiting, cheap enough to run all day long.

Model 5 min read Updated 2026-05-23
— What it is

Claude Sonnet 4.6, in plain words

Put simply: it's the Anthropic model you use day to day in claude.ai and Claude Code when you have no particular reason to pick a different one.

It does pretty much everything: write text, analyze a PDF, help me structure an email, read 100 pages of docs and hand me back a summary. It has a 200K-token context window (roughly 150K words), which means I can throw a whole big file at it in one go.

On the technical side, it's a "hybrid" model: it can answer quickly when the question is simple, or take the time to think when the question deserves it. I don't need to know which; it adapts on its own.

What I use it for

The cases where I reach for Sonnet 4.6 over anything else:

  • Writing tricky emails: a note to an important client, a message that has to be firm and diplomatic at the same time. Sonnet finds the right tone without overdoing it.
  • Analyzing a document: I drop in a contract, a market study, a report, and ask what it thinks. It flags the things that are actually off.
  • Thinking out loud: I describe a business problem, it asks me questions, and we dig into it together. Better than brainstorming solo.
  • Prepping a meeting: "Here's the topic, here are the attendees, what did we miss?" It comes back with 5 angles I wouldn't have thought of.
  • Coding via Claude Code: when I ask Claude to whip up a script for me, Sonnet 4.6 is what answers by default. Plenty good for 95% of the stuff.

How it compares

vs Claude Opus 4.7: Opus is smarter on the genuinely gnarly questions (complex reasoning, delicate code), but it's ~5× pricier and slower. I only switch to Opus when Sonnet gets stuck or when the stakes are high.

vs Claude Haiku 4.5: Haiku is ~3× cheaper and faster, but you lose some finesse. I use it for high-volume tasks (sorting 1,000 emails) rather than for conversations.

vs GPT-5: GPT-5 is slightly better at pure math reasoning. Sonnet 4.6 is better at writing, at code, and at tone. It's a tie on document analysis.

What it costs

API pricing as of May 23, 2026 (source: anthropic.com/pricing):

  • Input: $3 / million tokens
  • Output: $15 / million tokens
  • Prompt caching: up to 90% savings on repeated inputs

Using it via claude.ai (the web app): $20/month for the Pro plan, reasonable unlimited access. The Max plan at $100-200/month gives you a lot more volume if you run Claude Code nonstop.

My take

It's my default model. Full stop. When I don't know what to use, I go with Sonnet 4.6 and I rarely regret it.

What I love: the tone, the rigor on numbers, the fact that it says "I don't know" when it doesn't know. What sometimes bugs me: it can be too cautious on topics that are actually harmless.

If you're new to AI and you want a single subscription, this is the one.

Quick questions

Sonnet 4.6 or Opus 4.7, which should I pick?

Claude Sonnet 4.6 is Anthropic's balanced model, sitting between Haiku and Opus in the lineup. Released in early 2026, it's built for everyday use: writing, document analysis, code, with a 200,000-token context window.

Is Sonnet 4.6 multimodal?

Yes, it reads images, PDFs, diagrams. Not video natively yet.

How many tokens in its context?

200,000 tokens, or about 150,000 words, or 300-400 pages.

Verified 2026-05-25 · next review 2026-11-25

Pricing and specs verified on Anthropic's official pricing page. My own usage: ~3-4h/day on Sonnet 4.6 via Claude Code and claude.ai since it launched.

Want to keep going in order?

Head back to the full glossary, search a word, then open only the pages that deserve more than a short definition.

Open the AI glossary