Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.
Supported API features and tools
Text and image input; text output
Adaptive thinking with configurable effort
Messages API tool-use workflows with an appropriate integration
Prompt caching with 5-minute and 1-hour write options
Asynchronous Message Batches processing
Before you choose
No native audio, video, or image output; accepting images does not make this an image generator.
Forced tool selection is not supported; do not carry over a tool_choice configuration that forces a particular tool.
Non-default temperature, top_p, or top_k values are rejected.
Thinking blocks are tied to the generating model and conversation; changing models or editing earlier turns can invalidate them.
Adaptive thinking and effort
lowmediumhigh · defaultxhighmax
Adaptive thinking is on by default, with high effort as the API default. Choose low, medium, high, xhigh, or max based on measured results. To turn off up-front thinking, use between_tools at high or below; disabled is rejected, and between_tools is rejected at xhigh/max. Allow room in max_tokens for thinking plus the visible answer.
Anthropic lists this model as active; released September 28, 2026. Retirement is not sooner than September 28, 2027. This is an availability commitment, not a scheduled retirement.
Normal Messages output is limited to 128,000 tokens. Up to 300,000 output tokens are available only through the Message Batches API beta with the output-300k-2026-03-24 header.
The displayed cutoff is Anthropic’s reliable knowledge cutoff, published to month precision. A cutoff does not replace checking current facts or supplied evidence.
Effort levels are recalibrated from Sonnet 5: reevaluate quality and latency instead of transferring a setting unchanged.
Text between tool calls is returned in thinking blocks. An integration must handle the documented display settings if it wants visible progress updates.
The earlier computer_20251124 tool is not accepted on the Claude API and Google Cloud; consult the migration guide before reusing a computer-use integration.
Put it to work
Start with a more useful prompt.
Original examples from EZ Ai Assist. Adapt these to your task and the features available in your workspace.
Workflow 01
Review a small feature against its contract
Tie proposed edits to explicit requirements and observable behavior.
Review this feature implementation against the acceptance criteria and current code below. Identify missing cases, interface mismatches, and unintended behavior changes. Cite the requirement and code location for each finding. Suggest the smallest corrective patch and the tests needed to verify it. Keep unrelated refactoring out of scope and mark untested assumptions clearly.
Workflow 02
Compare options from a document packet
Make recommendations traceable to supplied evidence.
Compare the three options in these documents using the evaluation criteria below. For each criterion, record supporting evidence with a document and section reference, uncertainty, and any conflicting statement. Recommend an option only if the evidence supports one; otherwise identify the smallest missing information needed for a decision. Do not use unstated capabilities or costs.
Workflow 03
Translate meeting notes into accountable work
Separate agreed actions from ideas still under discussion.
Turn these meeting notes into an action register with task, named owner, due date, dependency, and source excerpt. Use unassigned or not stated when the notes omit a value. Separate agreed actions from suggestions and open questions. Highlight conflicting deadlines without resolving them yourself, and finish with the clarification questions the team should answer before work begins.
Developer reference
Anthropic API pricing
These are Anthropic API reference prices, not EZ Ai Assist subscription prices.
Standard rates apply across the full 1M-token context window; no separate long-context surcharge is listed for this model.
Batch processing discounts input and output by 50%. Cache writes and reads have distinct rates; calculate from actual usage rather than applying a cache-read discount to the entire request.
Thinking tokens are billed as output even when their text is hidden. Tool charges and optional US-only inference (1.1× token pricing) can also affect the total.
Its standard rates are $2 input and $10 output per million tokens. Do not apply Opus Fast-mode pricing to Sonnet.
Verify current provider pricing before estimating spend. These rates do not describe an EZ Ai Assist subscription.
Common questions
A few things worth knowing.
What changes when moving from Sonnet 5?
The migration includes different thinking behavior, restrictions on forced tools, and changes to thinking-block reuse. Effort levels are recalibrated. Test your existing prompts and integration rather than assuming the model-ID replacement alone preserves behavior.
Can I disable thinking?
The disabled setting is rejected. Use between_tools to remove up-front thinking at low, medium, or high effort; adaptive thinking is required at xhigh or max. Between-tools behavior and visible thinking summaries are separate concerns.
Why might streamed progress disappear?
Inter-tool text is returned in thinking blocks. If an integration renders only ordinary text, it may show no progress between tool calls. Follow the model’s display guidance instead of treating a quiet UI as proof that processing stopped.
Can it force a particular tool call?
The model does not support forced tool use. Review tool_choice behavior during migration and let the supported configuration select tools; do not assume a previous model’s forced-tool setup remains valid.
Can it return 300,000 tokens?
The normal Messages limit is 128,000 output tokens. A separate Message Batches beta permits up to 300,000 with the documented output-300k-2026-03-24 header. That beta is not the ordinary response limit or an app entitlement.
Is September 28, 2027 a shutdown date?
No. Anthropic’s commitment is that retirement will not occur sooner than that date. Sonnet 5.5 is currently active; a not-before commitment is not a scheduled retirement announcement.
Are tools and API rates part of my subscription?
No. This guide describes Anthropic’s API. EZ Ai Assist plans and enabled features are separate, and a provider capability does not automatically create a control in the app. Review output before acting on it.
Check the source
Official documentation
Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.