Chat

Model developer guide

Anthropic

15 website model guides

Use Claude for work that benefits from a clear brief, careful reasoning and an explicit review step. Compare Sonnet and Opus on your own writing or coding task, then check the exact version’s thinking and tool contract before treating an API feature as an app capability.

Long contextWritingResearch

Practical guidance

Where should I start?

Editorial starting points, not a performance ranking. Check current app access before choosing a model.

A balanced writing or development task

Evaluate Sonnet 5.5 for a task with a defined deliverable: a revised document, a code-review explanation or a plan supported by supplied material. Anthropic positions this version around the combination of speed and intelligence.

Tradeoff: That positioning is not a measured app benchmark. Compare factual accuracy, time to a useful answer and the amount of editing your own task needs.

Claude Sonnet 5.5 guide

A longer, evidence-heavy problem

Consider Opus 5.5 for a multi-stage coding or knowledge-work brief where constraints interact. Its official positioning is long-running agentic coding and knowledge work; structure the assignment into reviewable checkpoints.

Tradeoff: A long-running model is not automatically an autonomous app workflow. Tool execution, approvals and successful completion still depend on the integration.

Claude Opus 5.5 guide

Technical details

What does the provider document?

Official documentation describes the named model or gateway—not every feature available in this app.

Context and output are different budgets

Sonnet 5.5 and Opus 5.5 each publish a 1M-token context window and a 128K-token maximum output for synchronous Messages API requests. Those are endpoint-specific limits, not an app upload allowance or a ceiling for every beta service. Conversation history and thinking consume budget too, so a large source bundle can leave less room for the useful answer.

Thinking is not one universal on/off switch

Both 5.5 models use adaptive thinking by default and reject thinking.type disabled. Sonnet 5.5 has a between_tools mode that removes up-front thinking at high effort or below; Opus 5.5 does not accept that mode. Earlier Claude versions follow different rules, so a configuration copied from an old example can fail.

Complete website coverage

Explore all 15 model guides

Every existing catalog entry for this provider is linked below. Status labels come from the individual website guides; catalog coverage is not a guarantee of current app access. Historical and snapshot pages remain references, not extra active models.

Practical guidance

From a brief to a useful result

Original prompt examples, not generated-output demonstrations. They have not been generation-tested.

Workflow 1

Repair a confusing customer explanation

Provide the original explanation, verified product facts and the intended reader. Exclude private customer information. Ask for one small revision before a complete rewrite.

Rewrite the explanation below for a customer who understands the problem but not our terminology. Preserve every verified limitation and price condition. Do not invent features, guarantees or timelines. Return a clear short version, a list of ambiguous statements you could not resolve, and a change log showing what you simplified and why.

Check the result: Compare the revision against your approved facts. Confirm that friendlier wording did not remove an important limitation or create a new promise.

Start with Sonnet 5.5

Workflow 2

Review a risky software change before implementation

Supply a narrow diff, surrounding interfaces and expected behavior. State that the task is review only; pasted code does not give the model access to a repository.

Review this proposed change for behavior regressions. Trace each concern to a specific supplied line or interface. For every actionable finding, explain the failing scenario and the smallest test that would expose it. Separate confirmed defects from questions requiring more code. Do not claim you executed tests, and do not rewrite unrelated sections.

Check the result: Reproduce important findings locally, dismiss claims unsupported by the supplied code, and keep human approval before modifying or merging the change.

Use the Opus 5.5 guide

Technical details

What changes inside EZ Ai Assist?

Official Anthropic model references and thinking documentation support the provider details below. The website catalog establishes the linked guides, not current app access. Chat request parameters, tools and account settings were not independently inspected or generation-tested.

Website catalog

What does this provider page establish?

The full Claude guide list preserves the website’s existing identities, including older references. These pages help compare versions; they do not establish which models your account can currently select or whether a provider retirement has been handled in the app.

Unverified chat controls

What is exposed in EZ Ai Assist?

The current reasoning selector, accepted attachment types, maximum upload size and tool interface were not inspected for this provider update. Use the controls actually visible in your app session; do not infer computer use or repository access from Anthropic documentation.

Unverified upstream request

What reaches Anthropic?

The selected API model ID, thinking mode, effort, max_tokens, message transformation and tool-choice fields were not captured. The precise handling of 5.5 breaking changes remains an application-level verification task, not something this website-only update changes.

Before you rely on a result

Limits, costs and tradeoffs

  • Adaptive thinking and a larger context window do not remove factual errors. Require references to supplied evidence and verify calculations, quotations and code behavior.
  • Claude on Anthropic’s own products, the direct API and EZ Ai Assist are separate product experiences. Product subscriptions, tools, usage limits and data settings do not automatically transfer between them.
  • Longer reasoning can change latency and token use. The provider’s API rates are not EZ Ai Assist customer charges, and this guide does not infer a cost per conversation.

Provider references and picker estimates are not EZ Ai Assist customer charges. Read the website pricing explanation.

Compare by task

Other paths to explore

OpenAI

Compare the same document or code review with another reasoning family, using identical acceptance criteria.

Mistral AI

Explore a different range of text-and-vision models for bounded extraction or development tasks.

Kimi

Compare long-context coding and evidence organization, while checking different thinking and input contracts.

Common questions

Anthropic, without the assumptions

Should every difficult prompt go to Opus?

Not automatically. First make the task and evidence clear. Evaluate Sonnet and Opus on representative cases, then choose based on correct output, review effort and response time rather than the family name alone.

Can I turn thinking off on Claude 5.5?

The cited API documentation rejects disabled for Sonnet 5.5 and Opus 5.5. Sonnet’s between_tools mode is a distinct option, not a general off switch. Whether the app offers any of these controls is unverified.

Why does a review mention tests without proving they passed?

A text answer can propose or reason about a test. A passing result requires actual execution and its output. This guide’s prompts deliberately keep suggested checks separate from completed verification.

New to the terminology?

Official documentation

Sources & review notes

Highlighted provider contracts reviewed . Individual model guides carry their own review dates. Sources can change; these pages do not perform live health checks.

Official Anthropic model references and thinking documentation support the provider details below. The website catalog establishes the linked guides, not current app access. Chat request parameters, tools and account settings were not independently inspected or generation-tested.