Claude Sonnet 5.5

Sonnet 5.5 is the Claude to leave on as your default: it handles daily coding, agents that call tools and ordinary business tasks, and you save Opus for the rare heavy job. It follows Claude Sonnet 5 in Anthropic's Sonnet line, keeps the same 1M-token window and 128K output, and on Kadraris its per-token rate is lower than Sonnet 5.

Adaptive thinking is on from the start. You decide how deep it goes with reasoning_effort, which defaults to high.

What you can use it for

• The standing model in Claude Code, Cursor or Cline
• Agents that work through APIs and tools over many steps
• Writing tests, reviewing code and fixing bugs in big repositories
• Long contracts, logs or specs read in one request

Settings and limits

• reasoning_effort: drop to low or medium for simple tasks, because thinking is billed like output.
• thinking: {"type": "disabled"} returns a 400 error. Lower the effort instead.
• Through /v1/messages you can send thinking: {"type": "between_tools"} at high effort or lower.
• Forcing a tool with tool_choice set to any or a named tool returns 400. Use auto and say which tool to use in the instruction.
• Custom temperature, top_p and top_k values are rejected, same as on Sonnet 5.

Example request

from openai import OpenAI

client = OpenAI(
    base_url="https://kadraris.com/v1",
    api_key="sk-kd-...",
)
resp = client.chat.completions.create(
    model="claude-sonnet-5.5",
    reasoning_effort="medium",
    messages=[{"role": "user",
               "content": "Read this stack trace and "
                          "suggest the smallest fix."}],
)
print(resp.choices[0].message.content)

Set reasoning_effort to low for quick edits and back to high for tricky bugs.

Questions

What does Sonnet 5.5 cost on Kadraris?

$1.89 / $9.45 per 1M tokens, charged per token from your balance. Reasoning tokens count as output, so effort level has a direct effect on the bill. Failed requests cost $0.

What do I need to change when coming from Sonnet 5?

Remove forced tool choice and any non-default sampling values, then re-test your effort level, since Anthropic recalibrated what each level means. The Sonnet 5.5 vs Sonnet 5 page shows both side by side.

Can I use it from the Anthropic SDK?

Yes. Call /v1/messages with your sk-kd- key. The Claude Code guide covers the CLI setup.

Claude Sonnet 5.5 in your code

from openai import OpenAI

client = OpenAI(base_url="https://kadraris.com/v1", api_key="sk-kd-...")
resp = client.chat.completions.create(
    model="claude-sonnet-5.5",
    messages=[{"role": "user", "content": "Suggest three opening shots for a short film about a cafe."}],
)
print(resp.choices[0].message.content)

curl https://kadraris.com/v1/chat/completions \
  -H "Authorization: Bearer $KADRARIS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "claude-sonnet-5.5", "messages": [{"role": "user", "content": "Suggest three opening shots for a short film about a cafe."}]}'

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://kadraris.com/v1", apiKey: process.env.KADRARIS_API_KEY });
const resp = await client.chat.completions.create({
  model: "claude-sonnet-5.5",
  messages: [{ role: "user", content: "Suggest three opening shots for a short film about a cafe." }],
});
console.log(resp.choices[0].message.content);

export ANTHROPIC_BASE_URL=https://kadraris.com
export ANTHROPIC_AUTH_TOKEN=sk-kd-...
export ANTHROPIC_MODEL=claude-sonnet-5.5
claude