Claude Sonnet 5.5: what changed, pricing, and when to use Opus instead

Key takeaways

• Claude Sonnet 5.5 launched September 28, 2026 as the second model in the Claude 5.5 family (after Opus 5.5).
• API list price matches Sonnet 5’s token rates: $2 input / $10 output per 1M, but Anthropic claims lower cost per task because it uses fewer tokens and runs 30%+ faster.
• Specs that matter: 1M context, 128K max output, adaptive thinking, id claude-sonnet-5-5.
• Use Sonnet 5.5 as the default. Keep Opus 5.5 for the hardest work.
• Product home for the assistant remains Claude; this page is the Sonnet 5.5 model note.

Table of contents

1. What Sonnet 5.5 is
2. Claude 5.5 family snapshot
3. API pricing
4. Sonnet 5.5 vs Sonnet 5 vs Opus 5.5
5. Claude.ai vs API
6. Who should move now
7. FAQ

---

What Sonnet 5.5 is

Anthropic ships several “Claude” surfaces at once. For builders, pin the model id.

Claude Sonnet 5.5 is Anthropic’s fast, cost-efficient 5.5-family model for coding and knowledge work. It keeps Sonnet 5’s per-token list price while aiming to finish typical tasks with fewer tokens and lower latency.

What that means in practice:

• What it is: the workhorse id you should point most agents and Claude Code sessions at first.
• What “30% cheaper” means: Anthropic’s claim is per task, not a lower $/1M sticker. Measure in your harness.
• What it is not: Fable/Mythos (different Claude lines; see Fable 5.1 / Mythos 5.1), and not a reason to ignore Opus on brutal tasks.

Last verified: October 2026 against Anthropic’s Sonnet 5.5 announcement, Claude platform docs, and our Claude tool page.

Claude 5.5 family snapshot

Model	Role	API in/out (per 1M)	Context / max out
Claude Sonnet 5.5	Default fast workhorse	$2 / $10	1M / 128K
Claude Opus 5.5	Stronger, costlier	$4 / $20	1M / 128K
Other 5.x lines (e.g. Fable)	Specialized / higher tiers	Higher	See product docs

If your team still says “just use Claude,” make them say which id. Billing arguments disappear once the picker is explicit.

API pricing

Meter	Sonnet 5.5	Opus 5.5
Input	$2 / 1M	$4 / 1M
Output	$10 / 1M	$20 / 1M
Cache read	$0.20 / 1M	$0.20 / 1M
Cache write (5m / 1h)	$2.50 / $4	$5 / $8
Batch	50% off in/out	50% off in/out
US-only inference	1.1× where offered	1.1× where offered

Consumer subscriptions (Free / Pro / Max) are separate from API meters. Pro is still the individual seat most people compare to ChatGPT Plus.

Sonnet 5.5 vs Sonnet 5 vs Opus 5.5

Question	Practical answer
Upgrade from Sonnet 5?	Yes for most production agents after a short eval. Same $/1M, better speed/efficiency claims.
Replace Opus 5.5 everywhere?	No. Keep Opus on tasks where Sonnet fails quality gates.
Best default in Claude Code?	Start Sonnet 5.5; escalate to Opus when the agent spins or quality drops.

A simple way to think about it: Sonnet 5.5 is the daily driver. Opus is the tow truck.

Claude.ai vs API

Surface	Why it matters
claude.ai / apps	Human chat, Projects, Research, artifacts; plan limits apply
Claude Code	Agentic coding harness on Pro+
Claude API / cloud MaaS	Metered claude-sonnet-5-5 in your own product

If you only use the website, “we switched the API id” changes nothing until the product exposes Sonnet 5.5 in your picker.

Who should move now

Move defaults to Sonnet 5.5 if:

• You already standardize on Sonnet-class models
• Latency and tokens/task dominate your bill
• Claude Code is a daily driver for the team

Stay cautious if:

• Your eval suite still prefers Opus (or Fable) on critical paths
• You have not regression-tested tool-use prompts after the between_tools thinking migration Anthropic called out for some Sonnet setups

Related: Claude, Claude Fable 5.1 and Mythos 5.1.

FAQ

What is Claude Sonnet 5.5?

Claude Sonnet 5.5 is Anthropic’s September 28, 2026 Sonnet model in the Claude 5.5 family. It is positioned as faster and more efficient than Sonnet 5 for coding and knowledge work, with a 1M-token context window.

How much does Claude Sonnet 5.5 cost?

On the Claude API, Sonnet 5.5 is $2 per 1M input tokens and $10 per 1M output tokens, with $0.20 per 1M for cache reads. Batch is 50% off. Consumer Claude Pro remains a separate subscription (about $17 to $20/mo).

Is Sonnet 5.5 better than Opus 5.5?

Not universally. Sonnet 5.5 is the default workhorse for speed and cost. Opus 5.5 costs more ($4/$20 per 1M) and is the pick when you need the stronger model in the family.

What is the model ID for Sonnet 5.5?

Use claude-sonnet-5-5 on the Claude API and matching IDs on Bedrock, Google Cloud, and Microsoft Foundry. Confirm the live catalog if your region uses a prefixed form.

Does Sonnet 5.5 replace Claude on claude.ai?

Anthropic made Sonnet 5.5 available across Claude surfaces including claude.ai. Exact default routing can still vary by plan and feature (chat vs Claude Code). Check the model picker in your account.