Claude Opus 5.5 and Sonnet 5.5 API: Pricing in TL, What Changed from Opus 5, How to Switch

Anthropic's current Opus- and Sonnet-class models, Claude Opus 5.5 and Claude Sonnet 5.5, are available in the Onysoft AI Gateway catalog as anthropic/claude-opus-5.5 and anthropic/claude-sonnet-5.5. Both offer a 1,000,000-token context window and a 128,000-token maximum output. On price there are two plain facts: Opus 5.5 lists 20% below Opus 5 ($6 per 1M input tokens, $30 per 1M output tokens), while Sonnet 5.5 costs the same as Sonnet 5 ($3 input, $15 output).

This guide covers where the two models sit in the catalog, a USD and TL price table, the half-price :batch IDs, the one-line migration from Opus 5 and Sonnet 5, and working Python and curl examples. Prices, context and output limits are taken from the Onysoft catalog as of October 5, 2026; every example runs against Onysoft's OpenAI-compatible API.

Where Opus 5.5 and Sonnet 5.5 Sit in the Catalog

The current Claude lineup has four classes side by side: Fable 5.1 in the top price tier, Opus below it, Sonnet for day-to-day production traffic, and Haiku at the lowest unit cost. Opus 5.5 and Sonnet 5.5 replace the previous release in their own class (Opus 5 and Sonnet 5); the previous releases remain active in the catalog.

• Claude Opus 5.5 (anthropic/claude-opus-5.5): per the maker's catalog description, a model for demanding reasoning, coding and long-horizon agentic work that succeeds Claude Opus 5; the description singles out multi-step changes in large codebases.
• Claude Sonnet 5.5 (anthropic/claude-sonnet-5.5): per the maker's description, the Sonnet-class model for well-scoped everyday work and a direct upgrade to Claude Sonnet 5; building features and fixing bugs are highlighted.

The technical frame is identical for both: a 1,000,000-token context window and a 128,000-token maximum output. Because these match Opus 5 and Sonnet 5 exactly, a migration needs no changes for prompt length or output limits.

We do not quote benchmark scores in this article: the fastest way to see how a model performs on your work is to run the same prompt against both versions in the Playground. For the whole family, see the Claude models page.

Claude Opus 5.5 and Sonnet 5.5 Pricing: USD and TL Table

The table below shows Onysoft list prices per 1 million tokens for the current Claude lineup:

Model	API ID	Context (tokens)	1M input	1M output	1M input (TL)	1M output (TL)
Claude Opus 5.5	anthropic/claude-opus-5.5	1,000,000	$6.00	$30.00	294.21 TL	1,471.04 TL
Claude Opus 5.5 (batch)	anthropic/claude-opus-5.5:batch	1,000,000	$3.00	$15.00	147.10 TL	735.52 TL
Claude Sonnet 5.5	anthropic/claude-sonnet-5.5	1,000,000	$3.00	$15.00	147.10 TL	735.52 TL
Claude Sonnet 5.5 (batch)	anthropic/claude-sonnet-5.5:batch	1,000,000	$1.50	$7.50	73.55 TL	367.76 TL
Claude Opus 5	anthropic/claude-opus-5	1,000,000	$7.50	$37.50	367.76 TL	1,838.81 TL
Claude Sonnet 5	anthropic/claude-sonnet-5	1,000,000	$3.00	$15.00	147.10 TL	735.52 TL
Claude Fable 5.1	anthropic/claude-fable-5.1	1,000,000	$15.00	$75.00	735.52 TL	3,677.61 TL
Claude Haiku 4.5	anthropic/claude-haiku-4.5	200,000	$1.50	$7.50	73.55 TL	367.76 TL

Prices are Onysoft list prices as of October 5, 2026; see /models for live pricing. TL equivalents use 1 USD = 49.0348 TL (Central Bank of the Republic of Türkiye rate for October 2, 2026); billing uses the current central bank rate at the time of the request.

Three concrete takeaways from the table:

• Opus 5.5 is 20% cheaper than Opus 5. $6 instead of $7.50 per 1M input tokens, $30 instead of $37.50 per 1M output tokens.
• Sonnet 5.5 costs the same as Sonnet 5. $3 per 1M input, $15 per 1M output; switching does not change your budget.
• The Opus 5.5 batch price equals the standard Sonnet 5.5 price. anthropic/claude-opus-5.5:batch is $3 per 1M input and $15 per 1M output.

Example monthly bill: an application that processes 10 million input tokens and 2 million output tokens a month pays $150 (7,355.22 TL) on Opus 5, $120 (5,884.18 TL) on Opus 5.5, and $60 (2,942.09 TL) on Sonnet 5.5. Use the cost calculator to run the numbers for your own volume.

Billing is based on input and output token counts; there is no subscription or monthly commitment. Every response returns the amount for that request in USD in the cost field, and the dashboard shows the TL equivalent.

The :batch IDs: When Does Half Price Make Sense?

Both models also have a separate ID with the :batch suffix, billed at half the standard list price:

• anthropic/claude-opus-5.5:batch — $3 per 1M input, $15 per 1M output
• anthropic/claude-sonnet-5.5:batch — $1.50 per 1M input, $7.50 per 1M output

A batch ID suits bulk work where the answer does not have to arrive immediately: overnight evaluation sets, labeling document archives, regression tests, bulk summarization. To use it, you only change the model name in the request. Keep the standard ID for flows where a user is waiting on screen, and measure response times with your own workload before putting a batch ID into production.

One practical consequence: a bulk job you run on Sonnet 5.5 can run on the Opus 5.5 batch ID for the same budget, so you can compare which one gives better results on a small sample.

Migrating from Opus 5 and Sonnet 5: One Line

The only thing that has to change in your code is the model name:

Old ID	New ID	Price effect	Context / output
anthropic/claude-opus-5	anthropic/claude-opus-5.5	20% lower	Same (1,000,000 / 128,000)
anthropic/claude-sonnet-5	anthropic/claude-sonnet-5.5	Unchanged	Same (1,000,000 / 128,000)

The order we suggest:

1. Pick a representative sample of your existing prompts and run both versions side by side in the Playground or in your own test script.
2. Record the usage and cost fields of each response; token consumption can differ between versions, so read cost from actual usage rather than from the list price.
3. Move traffic gradually. anthropic/claude-opus-5 and anthropic/claude-sonnet-5 remain active in the catalog, so rolling back is also a one-line change.

Aliases: if you would rather not update code on every release, use the ~anthropic/claude-opus-latest and ~anthropic/claude-sonnet-latest aliases (the leading ~ is required). They redirect to the latest model in the family; as of October 5, 2026 they are listed in the catalog at the same price as Opus 5.5 and Sonnet 5.5. If you do not want behavior to change on its own between releases in production, calling the fixed ID is the more predictable choice.

For the previous release see the Claude Opus 5 API guide, and for the top price tier of the current lineup see the Claude Fable 5.1 API guide.

Which Model for Which Job? Opus 5.5, Sonnet 5.5, Fable 5.1 and Haiku 4.5

Price tier and the nature of the job drive the choice. The mapping below is based on how the maker positions the models and on the unit prices in the table:

Workload	Where to start	Why
Multi-step changes in a large codebase, long agent tasks	anthropic/claude-opus-5.5	The area the maker highlights for Opus 5.5; 20% below the Opus 5 price.
Building features, fixing bugs, coding assistance, RAG answers	anthropic/claude-sonnet-5.5	The everyday work the maker describes for Sonnet 5.5; half the token price of Opus 5.5.
Non-urgent bulk analysis and labeling	anthropic/claude-opus-5.5:batch or anthropic/claude-sonnet-5.5:batch	Half the standard price, when response time is not critical.
Classification, short summaries, high-volume routine calls	anthropic/claude-haiku-4.5	Lowest unit price in the family; 200,000-token context.
The hardest tasks, where budget comes second	anthropic/claude-fable-5.1	Top price tier of the current lineup; 2.5 times the token price of Opus 5.5.

Rule of thumb: start with Sonnet 5.5, escalate the cases where quality falls short to Opus 5.5, and send bulk jobs to the batch IDs. If you want model choice made automatically per request on mixed traffic, OnyRouter (onysoft/auto) is an option; for the other cost levers see the AI API cost optimization guide.

First Request with Python and curl

To get started, create an account, generate an API key with the sk-ony- prefix in the dashboard, and add balance; no separate contract with the model maker is needed. Because the Onysoft endpoint follows the OpenAI schema, the official openai Python package works as is in streaming mode, and streaming responses arrive in OpenAI chunk format (in non-streaming responses the OpenAI object is in the data field):

from openai import OpenAI

client = OpenAI(
    base_url="https://api.onysoft.com/v1",
    api_key="sk-ony-YOUR_KEY",
)

stream = client.chat.completions.create(
    model="anthropic/claude-opus-5.5",
    messages=[
        {"role": "user", "content": "Split the functions in this module by responsibility and write a step-by-step refactoring plan."}
    ],
    stream=True,
)
for chunk in stream:
    if not chunk.choices:
        continue
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="")

A non-streaming Sonnet 5.5 request to the same endpoint with curl:

curl https://api.onysoft.com/v1/chat/completions \
  -H "Authorization: Bearer sk-ony-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-5.5",
    "messages": [{"role": "user", "content": "Summarize this error log and list the three most likely causes."}]
  }'

Non-streaming responses come back in a success/data envelope; the OpenAI-format response object is in the data field. To control thinking effort, send the reasoning parameter through extra_body in the Python SDK; it is passed through as is for models that support it. Higher effort can consume more tokens, which shows up in cost:

stream = client.chat.completions.create(
    model="anthropic/claude-opus-5.5",
    messages=[{"role": "user", "content": "Write a detailed plan for splitting this service into modules."}],
    extra_body={"reasoning": {"effort": "high"}},
    stream=True,
)

See the API documentation for supported request parameters and error codes, and the model catalog for all 748 models you can reach with the same key.