Access Anthropic Claude Opus 5.5—also searched as Opus 5.5— through EvoLink's unified chat API. Test long-context coding, tool use, prompt caching before integrating.
Anthropic·Text Generation·Available
From $3.800 / 1M input tokens$4.000 official price-5%
Always-on thinkingEffort controlTool usePrompt cachingWeb search toolChat + Messages API
Production routeThis rate reflects platform-side availability — only confirmed server errors (HTTP 500 / empty response) count as failures. User-side issues (content moderation, invalid params, cancellation) plus rate limits, timeouts and auth errors are excluded. Before real traffic arrives, empty buckets may display as available.Live
Live
Context
1M context · 128K max output
Best For
Long-running agentic coding, analyst-grade knowledge work, chart and screenshot reading
Anthropic’s newest Opus model for long-running agentic coding, knowledge work, and visual analysis, at a lower list price than Claude Opus 5. Five billing dimensions — input, cache write, cache read, output, and the web search tool — are metered separately.
Claude Opus 5.5
Anthropic’s newest Opus model
Selected
Model ID
claude-opus-5-5
Best for
Multi-step coding in a real repository, pull-request review, analyst-grade reports and financial models, and reading dense charts, diagrams, and screenshots — work where a better answer at a lower per-token price removes effort downstream.
Input
$3.800 / 1M-5%
258.4 cr / 1M$4.000official price
Cache write
$4.750 / 1M-5%
323 cr / 1M$5.000official price
Cache read
$0.192 / 1M-4%
13 cr / 1M$0.200official price
Output
$19.000 / 1M-5%
1292 cr / 1M$20.000official price
All rates are per 1M tokens, shown in USD and credits, and reflect your account's current pricing.
Claude Opus 5.5 pricing
Estimate what one Claude Opus 5.5 request costs before you integrate. The calculator uses your account's current rates, with official pricing as a reference.
Request calculator
Enter the token mix for one request and the number of successful tool calls.
Only successful server-side calls are billed per call; failed attempts have no tool fee, but tokens still apply.
Web search$0.010/ call·0.68 cr / call
Claude Opus 5.5 API for long-running coding agents and knowledge work
Call Anthropic’s newest Opus model through EvoLink’s unified API at $3.800 / $19.000 per million input / output tokens. Opus 5.5 succeeds Claude Opus 5 at a lower list price, keeps the 1,000,000-token context window and 128K max output, and runs with thinking always on — effort is the control for depth, latency, and cost. Live on EvoLink since its September 22, 2026 release; the migration notes and FAQ below cover the settings that change from Opus 5.
Claude Opus 5.5 is served on EvoLink under the model ID claude-opus-5-5 through Chat Completions · Anthropic Messages, with the same API key and balance you use for every other model. It offers a 1M context window and up to 128K output tokens, plus long-context coding, tool use, prompt caching.
Claude Opus 5.5 specs and capabilities
Numbers come from the EvoLink route configuration; capabilities are what the API exposes today.
Context window
1M tokens
Max output
128K tokens
Input
Text + images
Output
Text · JSON (structured output) · tool calls
Reasoning
Always-on reasoning
Tool use
Function calling with multi-step tool sequences
Prompt caching
Cache write + cache read rates
Server-side tools
Web search, billed per successful call
Protocols
Chat Completions · Anthropic Messages
Model ID
claude-opus-5-5
Where Claude Opus 5.5 is the right Opus route
Opus 5.5 costs less per token than Opus 5 and, in Anthropic’s testing, often finishes the same task with fewer tokens. Route to it by the value of a completed task, then tune effort before touching the prompt.
Multi-step coding in a real repository
Carry a change through a large codebase until the tests pass, review pull requests, and plan multi-file refactors. Anthropic reports that at the default medium effort it matched or beat Claude Opus 5 at high effort on such tasks, in fewer steps. Measure accepted patches, not snippet quality.
Analyst-grade knowledge work
Long analytical deliverables, financial models, and audits of large document sets. Anthropic reports it is much less likely than Opus 5 to state a figure or cite a source the inputs do not support, so check citations against your own evidence set.
Charts, diagrams, and screenshots
Read dense charts, diagrams, and UI screenshots where meaning depends on position — which boxes an arrow connects, or what changed between two versions. Anthropic reports higher visual accuracy than Opus 5 at every effort level, often without cropping tools.
When another route is the better choice
Everyday coding, high-volume agents, classification, and latency-sensitive chat usually belong on Claude Sonnet 5. Claude Fable 5.1 is the escalation for the hardest reasoning. Keep Claude Opus 5 only where a workload is already validated on it and you have not re-tested yet.
Two ways to use Claude Opus 5.5: EvoLink API or Agent
Use the EvoLink API for product backends and batch jobs, or call Claude Opus 5.5 from Codex, Claude, or Gemini for coding and analysis workflows. Both paths share the same EvoLink API key, balance, model ID, and request history.
Option 1
Integrate with the EvoLink API
Best for: product backends, batch jobs, automated pipelines
Send OpenAI-compatible Chat Completions (or Anthropic Messages) requests to EvoLink and control the model ID, system prompt, output budget, tools, and structured output.
1Create an EvoLink API key in the console
2Point your OpenAI or Anthropic SDK at the EvoLink base URL and select the model ID shown above
3Send one representative request and read the usage field for input, cached, and output tokens
4Set max_tokens and retries per task; keep tool-call IDs and results across turns
Best for: coding, review, and analysis tasks in Codex, Claude, and Gemini
Give the Agent the task, the inputs to include, and the acceptance criteria. It assembles the request, calls Claude Opus 5.5 through EvoLink, and returns the answer with token usage.
1Set EVOLINK_API_KEY in your local environment; never put it in code or a prompt
2Describe the task, the inputs to include, and the expected output format
3Ask the Agent to call Claude Opus 5.5 through EvoLink and show the request before sending
4Let the Agent report the answer, token usage, and any error body
Claude Opus 5.5 API code example and error handling
This example shows the shortest runnable request: an OpenAI-compatible Chat Completions call with a system prompt, a user message, and an output budget. Open the API tab for the complete parameter and response reference.
curl -X POST https://api.evolink.ai/v1/chat/completions \
-H "Authorization: Bearer $EVOLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-opus-5-5",
"messages": [
{ "role": "system", "content": "You are a senior engineer reviewing a pull request." },
{ "role": "user", "content": "Review this diff and list the blocking issues with file and line:\n<diff>" }
],
"max_tokens": 16000
}'
# Anthropic Messages is also available at /v1/messages with the same model
# ID.
# Thinking is always on for this model and counts toward max_tokens, so
# leave room for it. The response includes choices[0].message and a usage
# object (prompt_tokens, completion_tokens, and cache read / write tokens
# when caching applies).
Invalid request or unsupported parameter
Check the model ID, the messages array, and parameter ranges against the API reference; remove fields this route does not support.
Authentication or balance issue
Check the Authorization bearer token and confirm the available balance in the console.
Context length exceeded
Prompt tokens exceed the Claude Opus 5.5 context window. Trim or retrieve only the relevant evidence and reuse cached prefixes.
Rate limited (429)
Back off and retry with jitter; batch or queue requests instead of sending parallel bursts.
Content or tool call rejected
Review sensitive content, malformed tool-call arguments, and JSON schema mismatches before retrying.
Move from Claude Opus 5 to Opus 5.5 without replacing your stack
Keep one EvoLink client and API key. Most requests carry over unchanged; the four steps below cover the settings that behave differently on Opus 5.5.
01
Switch the model ID
Point the client at EvoLink, keep the same API key, and select claude-opus-5-5. Both the OpenAI-compatible Chat Completions endpoint and the Anthropic Messages endpoint accept it.
Low migration
02
Remove thinking-off settings
Thinking cannot be switched off on Opus 5.5 at any effort level. Drop thinking: disabled and budget_tokens, omit the thinking field or send adaptive, and read response content by block type rather than position.
Breaking change
03
Set effort explicitly
The default effort is medium, one level below Claude Opus 5’s high. Start at medium, test low and high on your own tasks, and lower effort before adding “think less” instructions to the prompt.
Tune
04
Route tasks by completed-task value
Compare success rate, retries, output tokens, cache hits, wall-clock time, and human correction against Claude Opus 5 and Sonnet 5 on replayed production samples — not token price alone.
Evaluate
Prompt caching for repeated agent context
Cache stable system prompts and tool schemas to reuse context across agent turns. Writes and reads are billed separately. Anthropic lists cache reads at 5% of its input rate; use EvoLink’s displayed write and read prices to budget for cache hits and misses.
1M-token context with a 128K max output
Keep connected code, specifications, documents, and agent state in one working context. Thinking counts toward max_tokens, so leave room for it: for long agentic coding turns, 64K is a reasonable starting point.
Thinking is always on — effort is the only control
Opus 5.5 always uses adaptive thinking, with medium effort by default. Anthropic rejects thinking: disabled and budget_tokens. EvoLink converts these legacy settings to adaptive thinking for compatibility; they cannot turn thinking off. Prefer omitting thinking or sending adaptive, and tune effort explicitly.
Forced tool choice is not supported
Forced tool choice is unsupported: Messages any/tool and Chat Completions required/named functions return an error. With auto, clearly request the tool and verify that a call occurred; auto does not guarantee one. Notes between tool calls may arrive as thinking blocks, so parse content by block type.
Claude Opus 5.5: five billing dimensions, metered separately
Claude Opus 5.5 does not bill as a single blended rate. Knowing which dimension a workload consumes — and that thinking is billed as output — is what keeps an Opus budget predictable.
Input and output tokens
$3.800 per 1M input tokens and $19.000 per 1M output tokens on EvoLink, compared with Anthropic’s $4.000 and $20.000. Thinking always runs and is billed as output even when its text is not returned, so size max_tokens for the thinking plus the reply.
5-minute cache writes and cache reads
Storing a prefix for 5 minutes costs $4.750 per 1M tokens; reusing it costs $0.192 per 1M tokens. Anthropic’s official cache-read rate is 5% of its input rate. EvoLink rates are rounded, so use the displayed prices when estimating savings.
Web search tool
Billed per search at $0.010. Search results also enter the request as ordinary input tokens, so a research-heavy turn pays both the per-search fee and the tokens it pulls in.
Compare Claude routes after workload testing
EvoLink
First check whether Opus 5.5 reduces retries and review effort on your own tasks at the effort level you plan to run. Then compare price, context, caching, and workload fit to choose the production route.
Multi-step coding in a real repository, pull-request review, analyst-grade reports and financial models, and reading dense charts, diagrams, and screenshots — work where a better answer at a lower per-token price removes effort downstream.
The previous Opus flagship at a higher list price; keep it where a workload is already validated and not yet re-tested.
Opus-class coding and agent performance at a balanced Sonnet rate for everyday production traffic.
Claude Opus 5.5
Input / output$3.8 / $19
Context1M
CachingRead + write
Best forMulti-step coding in a real repository, pull-request review, analyst-grade reports and financial models, and reading dense charts, diagrams, and screenshots — work where a better answer at a lower per-token price removes effort downstream.
First requests, effort, and the production checklist for the previous Opus model — most of it carries over, except that Opus 5.5 cannot disable thinking.
Is the Claude Opus 5.5 API available through EvoLink?
Yes. Claude Opus 5.5 is available on EvoLink as a production route under model ID "claude-opus-5-5".
When was Claude Opus 5.5 released, and is it the same model as “Opus 5.2”?
Anthropic released Claude Opus 5.5 on September 22, 2026, and it has been callable on EvoLink since launch day. No model named “Opus 5.2” was ever announced; that label came from community routing reports, and Anthropic has not said whether they described early Opus 5.5 traffic, so treat the two as separate names.
What model ID should I use for the Claude Opus 5.5 API?
Send model "claude-opus-5-5". The same ID works for both the OpenAI-compatible Chat Completions endpoint and the Anthropic Messages endpoint.
How much does the Claude Opus 5.5 API cost on EvoLink?
EvoLink pricing: $3.800 per 1M input tokens and $19.000 per 1M output tokens, against Anthropic’s $4.000 and $20.000.
Is Claude Opus 5.5 cheaper than Claude Opus 5?
Per token, yes: Anthropic lists Opus 5.5 about 20% below Opus 5 on input and output and about 60% below it on cache reads. Anthropic’s announcement also claims it costs about 40% less to run because it finishes tasks with fewer tokens; that is Anthropic’s measurement on its own task mix, so compare cost per accepted task on your own workload at the effort you plan to run.
How is Claude Opus 5.5 prompt caching billed?
5-minute cache writes cost $4.750 per 1M tokens and cache reads cost $0.192 on EvoLink, compared with Anthropic’s $5.000 and $0.200. They are metered separately. Use the displayed rates, including rounding, to estimate your workload.
How is the Claude Opus 5.5 web search tool billed?
Web search is charged per search at $0.010. Tokens the search results add to the request are billed normally.
Can I turn thinking off on Claude Opus 5.5?
No. Thinking is always on. Anthropic rejects thinking: disabled and budget_tokens; EvoLink converts legacy settings to adaptive thinking without disabling it. Omit thinking or send adaptive, and use effort (low, medium, high, xhigh, max; default medium) to control depth, latency, and cost.
What changes when I migrate from Claude Opus 5?
Change the model ID, remove thinking-off settings, and set effort explicitly: the default changes from high to medium. Forced tool choice is unsupported in both protocols (Messages any/tool; Chat Completions required/named functions). Use auto with strict tool schemas or structured outputs, and validate that a tool call happened. Read response blocks by type; context and max output are unchanged.
Should I switch from Claude Opus 5 or Claude Fable 5.1 to Opus 5.5?
From Opus 5: pilot Opus 5.5 on long agentic coding, large-document analysis, and cache-heavy agents, and keep Opus 5 as the rollback route until a matched replay at your chosen effort shows a lower cost per accepted task. From Fable 5.1: Anthropic says Opus 5.5 performs at Fable 5.1 level on most work at a fraction of the price, so re-test Fable workloads and keep Fable 5.1 only where higher-effort Opus 5.5 still falls short.
Which protocols and SDKs work with Claude Opus 5.5?
Use the OpenAI-compatible Chat Completions endpoint or the Anthropic Messages endpoint with the same EvoLink API key. Claude Code and other Anthropic-native clients work by pointing the base URL at EvoLink.
Do Claude Code subscription usage limits apply to the Claude Opus 5.5 API on EvoLink?
No. Claude Code and Claude app plans meter a separate subscription quota with its own reset window. Requests through EvoLink are billed per token against your EvoLink balance at the rates on this page, with no plan quota. Point Claude Code’s base URL at EvoLink to use the same claude-opus-5-5 route with your EvoLink key.
What context window and max output does Claude Opus 5.5 support?
The EvoLink route records a 1,000,000-token context window with up to 128K output tokens per response. Thinking counts toward max_tokens.
Why does my streaming UI go quiet between tool calls on Opus 5.5?
On Opus 5.5 the short notes the model writes between tool calls arrive as progress-update thinking blocks instead of text blocks, and their text is empty at the default display setting. Set thinking.display to “updates” (beta, header thinking-display-updates-2026-08-18) or “summarized”, render each non-empty thinking block before the tool call it precedes, and pass the blocks back unchanged.
What should a production Claude Opus 5.5 evaluation measure?
Track first-pass success, accepted deliverables, retries, output tokens at your chosen effort, cache hit rate, valid tool calls, time to accepted result, human correction, and fallback rate — not the per-token rate alone.
