How does Opus 5.5 differ from Opus 5?

On capability, Anthropic calls 5.5 a major step up from Opus 5, with output more than 30% faster and better prompt-injection resistance. On the API, thinking cannot be disabled, forced tool use returns a 400, the default effort drops from high to medium, thinking blocks are bound to the conversation, computer use on the Claude API accepts only computer_toolset_20260801, and a biology safety classifier is added. Context, output limit and tokenizer are unchanged.

When should I still use Opus 5 instead of 5.5?

When an integration cannot yet drop disabling thinking, forced tool_choice or computer_20251124, or it edits conversation history and replays thinking blocks. Anthropic commits to retiring Opus 5 no sooner than July 24, 2027.

How do I choose between Opus 5.5 and Fable 5.1?

Anthropic's advice is to start on Opus 5.5 and move to Fable 5.1 for demanding reasoning and long-horizon agentic work, or when evals on Opus 5.5 at higher effort still fall short. Both have a 1M context, 128K output and always-on thinking; the models overview rates Opus 5.5's comparative latency as Moderate and Fable 5.1's as Slower.

Claude Opus 5.5

Save 72%Official

anthropic/claude-opus-5-5

Anthropic's September 2026 Opus — successor to Opus 5, with cache reads at 5%.

Get an API key View code sample

Live model playground

Try Claude Opus 5.5 here

Tune the request, see the current price ceiling, and inspect the real response without leaving this model page.

streamchat

1Configure input

Prompt

163 / 20,000

Add images

Maximum estimated cost—

2Live resultExample

Claude Opus 5.5

B is telling the truth.

Quick check of each case, since only one can be truthful:

If A were truthful, then B lies, which makes "C is lying" false, so C is truthful too — two truth-tellers, contradiction.
If C were truthful, then A and B both lie. But if A lies, then B is actually truthful — contradiction with B lying.
If B is truthful, then C lies. C's claim "A and B are both lying" is false, which holds since B is truthful. A says "B is lying," which is false, so A lies. Only B tells the truth — consistent.

So the answer is B.

Context window: 1M
Max output: 128K
Input modalities: TextImage
Output modalities: Text
Capabilities: Vision · Reasoning · Tool calling · Coding · Prompt cache · Paid cache writes
Upstream release: 2026-09-22
RouteMux name: anthropic/claude-opus-5-5

What is Claude Opus 5.5

Claude Opus 5.5 was released by Anthropic on September 22, 2026, succeeding Opus 5, as the first model in the Claude 5.5 family, with retirement not sooner than September 22, 2027. Anthropic's models overview advises starting most workloads on Opus 5.5 when unsure, and moving to Fable 5.1 for demanding reasoning and long-horizon agentic work, or when evals on Opus 5.5 at higher effort still fall short. Anthropic says it performs at the level of Fable 5.1 on most work, and generates output more than 30% faster than Opus 5.

It has a 1M-token context window, up to 128K output tokens per request, image input and tool calling, and a reliable knowledge cutoff of June 2026. Adaptive thinking is always on and cannot be turned off; effort is the only thinking control, with all five levels and medium as the default, one level below Opus 5. At the same level it tends to think more than Opus 5, so effort settings carried over from older models need a fresh sweep.

Coming from Opus 5, Anthropic lists four breaking changes: thinking "disabled" or a manual budget returns a 400; forced tool use (tool_choice "any" or a named tool) returns a 400; thinking blocks are bound to the model and conversation, so replaying them after editing history can fail; and computer use on the Claude API accepts only computer_toolset_20260801. One quieter change: progress text between tool calls now arrives in thinking blocks that are empty by default. Anthropic rejects non-default temperature, top_p and top_k on this model; RouteMux strips them before forwarding.

Through RouteMux, Opus 5.5 costs $1.12 per 1M input tokens against Anthropic's $4, $5.6 per 1M output tokens against $20, and $0.056 per 1M cache-read tokens against $0.2.

Before you choose

How does Opus 5.5 differ from Opus 5?: On capability, Anthropic calls 5.5 a major step up from Opus 5, with output more than 30% faster and better prompt-injection resistance. On the API, thinking cannot be disabled, forced tool use returns a 400, the default effort drops from high to medium, thinking blocks are bound to the conversation, computer use on the Claude API accepts only computer_toolset_20260801, and a biology safety classifier is added. Context, output limit and tokenizer are unchanged.
When should I still use Opus 5 instead of 5.5?: When an integration cannot yet drop disabling thinking, forced tool_choice or computer_20251124, or it edits conversation history and replays thinking blocks. Anthropic commits to retiring Opus 5 no sooner than July 24, 2027.
How do I choose between Opus 5.5 and Fable 5.1?: Anthropic's advice is to start on Opus 5.5 and move to Fable 5.1 for demanding reasoning and long-horizon agentic work, or when evals on Opus 5.5 at higher effort still fall short. Both have a 1M context, 128K output and always-on thinking; the models overview rates Opus 5.5's comparative latency as Moderate and Fable 5.1's as Slower.

Claude Opus 5.5 API pricing

Prices verified 2026-10-06

InputOfficial $4 / M tokens

$1.12/ M tokensSave 72%

OutputOfficial $20 / M tokens

$5.6/ M tokensSave 72%

Cache read $0.056/M · Cache write $1.4/M

Estimate cost

Input tokensOutput tokens

Estimated cost$0.0168

Key capabilities

Anthropic's suggested starting point: The models overview (October 2026) advises starting most workloads on Opus 5.5 and moving to Fable 5.1 only when evals at higher effort still fall short.
Thinking is always on: It cannot be disabled and budget_tokens is not accepted. Effort sets the depth, from low to max with medium as the default, and per-message effort (beta) changes it without breaking the prompt cache.
Cache reads at 0.05x base input: Anthropic's pricing footnote bills cache hits on Opus 5.5 at 0.05 times the base input price, and the minimum cacheable prompt is 512 tokens. A long system prompt reused across a session is where this pays off.
Sharper reading of charts and screenshots: Anthropic says it reads dense charts and layout-dependent visuals much more precisely without tools, so vision workarounds written for earlier models may no longer be needed.

Limited time10% extra Credits on your first top-up — up to 200 bonus Credits

Applies to your first top-up within 72 hours of signing up. Minimum $20, once per account.

Interface and protocols

Anthropic MessagesStreaming

Use any of these IDs to call this model via the API.

RouteMux nameanthropic/claude-opus-5-5

Replace the ROUTEMUX_KEY placeholder with your API key. Create one →

from anthropic import Anthropic

client = Anthropic(
    base_url="https://api.routemux.com/anthropic",
    api_key="<ROUTEMUX_KEY>",
)

message = client.messages.create(
    model="anthropic/claude-opus-5-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "What is the meaning of life?"}],
)
print(message.content[0].text)

base_url: https://api.routemux.com/anthropic

Performance (last 30 days)

57,305 requests · last 30 days

Total usage13212.43M

P50 first-token latency3.52s

Avg throughput58 tok/s

Success rate100.0%

Uptime · 24htested via anthropic_messages

100.00%

Success rate · 60d · from real requests

Claude Opus 5.5 FAQ

›What is Claude Opus 5.5?

Anthropic's September 2026 Opus — successor to Opus 5, with cache reads at 5%.

›How large is the context window of Claude Opus 5.5?

Claude Opus 5.5 supports a context window of up to 1M tokens, with up to 128K output tokens per request.

›How much does the Claude Opus 5.5 API cost?

Through RouteMux, Claude Opus 5.5 costs $1.12 for input and $5.6 for output (USD per 1M tokens), below the official list price.

›How do I call Claude Opus 5.5 via API?

Claude Opus 5.5 is OpenAI-compatible: point your base URL at RouteMux and set the model field to anthropic/claude-opus-5-5 — no code changes needed.

›Which input and output modalities does Claude Opus 5.5 support?

Claude Opus 5.5 accepts text, image as input and produces text as output.

›When was Claude Opus 5.5 released?

Claude Opus 5.5 was released on September 22, 2026.

Sources

Reviewed by the RouteMux team on 2026-10-24

View as Markdown