Claude Opus 5.5 vs Sonnet 5.5 vs Haiku 5.5: Which Claude to Use

Anthropic shipped the whole Claude 5.5 family in about two weeks: Opus 5.5 on September 22, 2026, Sonnet 5.5 on September 28 and Haiku 5.5 on October 7. All three run on the Claude apps, Claude Code, the Claude Platform, AWS, Google Cloud and Microsoft Azure.

Prices per million tokens

Haiku's price rises to $0.50 in and $2.50 out for prompts over 100,000 tokens.

Opus 5.5: the heavy lifter

Anthropic says Opus 5.5 leads its benchmarks in agentic coding, computer use and knowledge work, generates output more than 30% faster than Opus 5 and costs less: cache reads, which make up most of agentic coding costs, are 60% cheaper. Anthropic also raised the five-hour usage limits on Pro, Max, Team and seat-based Enterprise plans with this release, and offers a fast mode in Claude Code with up to 2.5x speed.

Sonnet 5.5: the default

Sonnet 5.5 keeps Sonnet 5's price but needs far fewer tokens for the same work, runs over 30% faster and scores 70.6% on Terminal-Bench 4.0, up from Sonnet 5's 10.3%. In Claude Code and the apps its default effort is Medium. For most coding sessions it is the sensible first choice.

Haiku 5.5: cheap and quick

Haiku 5.5 is Anthropic's fastest model and, on average, about 75% cheaper to run than Haiku 4.5. Anthropic recommends it for live customer support, browser use and as a subagent under Opus or Sonnet, and says the bigger two remain better for complex agentic coding.

Using Claude 5.5 in Claude Code or Cursor

If you want usage without a monthly plan, the Claude API $100 credit ($10.99, 30 days) and $500 credit work in any tool that takes a custom endpoint, and the 10M-token day pass ($2.99) covers a single heavy day. These run through a third-party gateway; the models it serves are listed on your order.

Frequently asked questions

Which Claude is best for coding?

Sonnet 5.5 for everyday sessions, Opus 5.5 for long refactors and agent runs, Haiku 5.5 for fast subagents.

Is Opus 5.5 Anthropic's top model?

It is the top of the 5.5 family. Anthropic says it performs at the level of Claude Fable 5.1 for most tasks while costing 40% less to run than Opus 5.