Claude Opus 5.5, Opus 5, Sonnet 5 and Fable 5.1 through one API key

The current Claude family is available through InferAll with the same key you use for every other model: Claude Opus 5.5, Claude Opus 5, Claude Sonnet 5 and Claude Fable 5.1, plus the Claude 4 models.

Model id	Anthropic list price per 1M tokens (input / output)
anthropic/claude-opus-5-5	$4 / $20
anthropic/claude-opus-5	$5 / $25
anthropic/claude-sonnet-5	$2 / $10
anthropic/claude-fable-5-1	$10 / $50

These are paid models, billed from your InferAll balance. The free trial covers the open NVIDIA models only, so add the $5 starter pack at /billing first.

What we measured (2026-09-26, through the live API)

Two short prompts per model on /v1/chat/completions and one on /v1/messages. Every call answered as the model named.

Model	Chat completions	Messages
Opus 5.5	1.9s, 1.7s	1.1s
Opus 5	1.8s, 1.6s	1.5s
Sonnet 5	1.8s, 1.9s	1.7s
Fable 5.1	3.6s, 3.6s	3.0s

These are short prompts; long outputs take longer.

---

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(
    base_url="https://api.inferall.ai/v1",
    api_key="ifu_your_key_here",  # inferall.ai/keys
)

response = client.chat.completions.create(
    model="anthropic/claude-sonnet-5",
    messages=[{"role": "user", "content": "Review this function for bugs: ..."}],
    max_tokens=2000,
)
print(response.choices[0].message.content)

Python (Anthropic SDK)

import anthropic

client = anthropic.Anthropic(
    base_url="https://api.inferall.ai",
    api_key="ifu_your_key_here",
)

message = client.messages.create(
    model="anthropic/claude-opus-5-5",
    max_tokens=2000,
    messages=[{"role": "user", "content": "Plan a database migration for ..."}],
)
print(message.content[0].text)

Claude Code

Once your account has credits, pick any Claude model (Claude Code's own default Claude models also work without the last export):

export ANTHROPIC_BASE_URL=https://api.inferall.ai
export ANTHROPIC_API_KEY=ifu_your_key_here
export ANTHROPIC_MODEL=anthropic/claude-sonnet-5
claude

On the free trial, set ANTHROPIC_MODEL=nvidia/nemotron-3-super-120b-a12b instead: trial accounts cannot use paid models, and the gateway refuses a Claude id rather than answering with a different model.

---

Things that behave differently on these models

• Sampling parameters are not accepted by the Claude 5 models. If you send temperature or top_p through our OpenAI-compatible endpoint, InferAll drops them instead of failing the request.
• Forced tool use is not accepted by Fable 5.1 and Opus 5.5. If you send tool_choice: "required" or a named function through /v1/chat/completions, InferAll sends auto plus an instruction to call the tool. In our tests both models called the tool; it is a strong instruction rather than a guarantee.
• /v1/models lists these ids exactly as you should send them (anthropic/claude-...), so tools that build a model menu from it, such as Claude Desktop in gateway mode, work without edits.

Get started

1. Create an account at inferall.ai/keys.
2. Add the $5 starter pack at /billing.
3. Call any model above with your key. The full list is on the models page.