Skip to content
InfercomInfercomInfercom Documentation

Features

Implement Anthropic-Compatible Features - Developer Guide

Use the Anthropic SDK and Messages API with the Infercom API. Drop-in compatible - just change the base URL and API key to switch to EU sovereign inference.

The Infercom API supports Anthropic’s Messages API format (/v1/messages), enabling you to use the Anthropic Python SDK and compatible tooling with Infercom’s models. This is useful for applications and frameworks built around the Anthropic API, such as Claude Code, LangChain’s Anthropic provider, or custom agentic workflows.

Terminal window
pip install anthropic

Set the base_url to Infercom’s API and provide your Infercom API key.

import anthropic
client = anthropic.Anthropic(
base_url="https://api.infercom.ai",
api_key="your-infercom-api-key"
)
import anthropic
client = anthropic.Anthropic(
base_url="https://api.infercom.ai",
api_key="your-infercom-api-key"
)
message = client.messages.create(
model="gpt-oss-120b",
max_tokens=1024,
messages=[
{"role": "user", "content": "What is the capital of Germany?"}
]
)
print(next(block.text for block in message.content if block.type == "text"))

gpt-oss-120b returns a thinking block before the text block, so select the text block by its type. content[0].text raises an AttributeError. See Reasoning (thinking blocks).

import anthropic
client = anthropic.Anthropic(
base_url="https://api.infercom.ai",
api_key="your-infercom-api-key"
)
with client.messages.stream(
model="gpt-oss-120b",
max_tokens=1024,
messages=[
{"role": "user", "content": "Write a haiku about AI."}
]
) as stream:
for text in stream.text_stream:
print(text, end="", flush=True)

Use the system parameter to provide instructions to the model.

message = client.messages.create(
model="gpt-oss-120b",
max_tokens=1024,
system="You are a helpful assistant that speaks like a pirate.",
messages=[
{"role": "user", "content": "Hello!"}
]
)
message = client.messages.create(
model="gpt-oss-120b",
max_tokens=1024,
messages=[
{"role": "user", "content": "My name is Thomas."},
{"role": "assistant", "content": "Nice to meet you, Thomas!"},
{"role": "user", "content": "What is my name?"}
]
)

The Anthropic Messages API supports tool use for models that have function calling capabilities.

import anthropic
client = anthropic.Anthropic(
base_url="https://api.infercom.ai",
api_key="your-infercom-api-key"
)
message = client.messages.create(
model="gpt-oss-120b",
max_tokens=200,
tools=[
{
"name": "get_weather",
"description": "Get the current weather in a given location",
"input_schema": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "The city and country, e.g. Munich, Germany"
}
},
"required": ["location"]
}
}
],
messages=[
{"role": "user", "content": "What's the weather in Munich?"}
]
)
print(message.content)

If the model decides to use the tool, the response will include a tool_use content block:

[
{
"type": "tool_use",
"id": "call_abc123",
"name": "get_weather",
"input": {"location": "Munich, Germany"}
}
]

After executing the tool, send the result back to continue the conversation:

message = client.messages.create(
model="gpt-oss-120b",
max_tokens=200,
tools=[
{
"name": "get_weather",
"description": "Get the current weather in a given location",
"input_schema": {
"type": "object",
"properties": {
"location": {"type": "string"}
},
"required": ["location"]
}
}
],
messages=[
{"role": "user", "content": "What's the weather in Munich?"},
{
"role": "assistant",
"content": [
{
"type": "tool_use",
"id": "call_abc123",
"name": "get_weather",
"input": {"location": "Munich, Germany"}
}
]
},
{
"role": "user",
"content": [
{
"type": "tool_result",
"tool_use_id": "call_abc123",
"content": "Sunny, 22°C"
}
]
}
]
)
print(next(block.text for block in message.content if block.type == "text"))

gemma-4-31B-it accepts images as base64 image blocks. A URL image source returns HTTP 400. gpt-oss-120b is text only and returns HTTP 400 for an image.

import base64
with open("sample.jpg", "rb") as image_file:
image_base64 = base64.b64encode(image_file.read()).decode("utf-8")
message = client.messages.create(
model="gemma-4-31B-it",
max_tokens=1024,
messages=[
{
"role": "user",
"content": [
{
"type": "image",
"source": {"type": "base64", "media_type": "image/jpeg", "data": image_base64}
},
{"type": "text", "text": "What is happening in this image?"}
]
}
]
)
print(next(block.text for block in message.content if block.type == "text"))

Supported formats and the token cost per image: Vision.

gpt-oss-120b and MiniMax-M3.1-Flash return their reasoning as a thinking block before the text block. They do this with or without the thinking parameter. gemma-4-31B-it returns no thinking block.

On gpt-oss-120b, budget_tokens selects one of three reasoning levels. It does not set an exact token limit.

budget_tokens Reasoning level
Up to 4,096 low
4,097 to 16,384 medium, the same as without the thinking parameter
16,385 and above high

thinking: {"type": "disabled"} does not turn reasoning off. Reasoning tokens count toward max_tokens: see Budgeting tokens for reasoning.

message = client.messages.create(
model="gpt-oss-120b",
max_tokens=4096,
thinking={"type": "enabled", "budget_tokens": 1024},
messages=[{"role": "user", "content": "What is 27 * 43?"}]
)
for block in message.content:
if block.type == "thinking":
print("Thinking:", block.thinking)
elif block.type == "text":
print("Answer:", block.text)
import anthropic
import asyncio
async def main():
client = anthropic.AsyncAnthropic(
base_url="https://api.infercom.ai",
api_key="your-infercom-api-key"
)
message = await client.messages.create(
model="gpt-oss-120b",
max_tokens=1024,
messages=[
{"role": "user", "content": "Hello!"}
]
)
print(next(block.text for block in message.content if block.type == "text"))
asyncio.run(main())

You can also call the Messages API directly with curl:

Terminal window
curl https://api.infercom.ai/v1/messages \
-H "Content-Type: application/json" \
-H "x-api-key: your-infercom-api-key" \
-H "anthropic-version: 2023-06-01" \
-d '{
"model": "gpt-oss-120b",
"max_tokens": 1024,
"messages": [
{"role": "user", "content": "Hello!"}
]
}'
Parameter Type Description
model string Required. The model to use (e.g., gpt-oss-120b)
messages array Required. Array of message objects with role and content
max_tokens integer Required. Maximum tokens to generate
system string System prompt for the model
temperature number Sampling temperature (0.0-2.0). No default, unlike the Anthropic API, which defaults to 1.0.
top_p number Nucleus sampling parameter
top_k integer Top-k sampling parameter
stop_sequences array Custom stop sequences
stream boolean Enable streaming responses
tools array Tool definitions for function calling
tool_choice object Control tool usage (auto, any, or specific tool)
thinking object Accepted. On gpt-oss-120b, budget_tokens selects the reasoning level. See Reasoning (thinking blocks)

The Anthropic Python SDK 2.x has no temperature, top_p or top_k arguments. Pass them in extra_body, for example extra_body={"temperature": 0.7}.

The following Anthropic-specific features are not supported:

  • Prompt caching (cache_control)
  • URL image sources (use base64, see Images)
  • PDF file inputs
  • Citations
  • Server-side tools (web search, code execution)
  • Batch API
  • repetition_penalty (returns HTTP 400 on this endpoint; use /v1/chat/completions)
Aspect Anthropic Infercom
Models Claude (Opus, Sonnet, Haiku) Open-weight models. EU sovereign: gpt-oss-120b, gemma-4-31B-it. Global Model Catalog (processed outside the EU): MiniMax-M3.1-Flash
Base URL https://api.anthropic.com https://api.infercom.ai
API key header x-api-key x-api-key (same)
Version header Required: anthropic-version Supported but optional
Error shape Anthropic error format Anthropic error format (see below)

/v1/messages returns errors in the Anthropic shape, not the OpenAI envelope used by Infercom’s other endpoints. The body carries a top-level "type": "error", and the error object holds message and type only:

{
"type": "error",
"error": {
"type": "not_found_error",
"message": "The model `nonexistent-model` does not exist or you do not have access to it."
},
"request_id": "req_6313f64378ad4a89b6005ab233a3f2a9"
}
Field Notes
type (top level) Always error
error.type authentication_error, invalid_request_error or not_found_error
error.message Human-readable explanation
error.param / error.code Not returned. The OpenAI-compatible endpoints provide these; /v1/messages does not
request_id Present on every status. Most carry a req_ prefix; the 401 does not

This shape is returned for every error status, including 400, 401, 404, 410 and 422. See API error codes for the shapes the other endpoints use.

When to use Anthropic vs OpenAI compatibility

Section titled “When to use Anthropic vs OpenAI compatibility”
Use case Recommended API
Existing Anthropic SDK code Anthropic Messages API (/v1/messages)
Claude Code, LangChain Anthropic Anthropic Messages API (/v1/messages)
OpenAI SDK code OpenAI Chat Completions API (/v1/chat/completions)
Agentic workflows, coding tools Responses API (/v1/responses)
New projects Any - all three APIs work with the same models