Reasoning LLM

Large

gpt-oss-120b

Overview

gpt‑oss‑120b is the most powerful model in the gpt‑oss family. With 117 billion total parameters and 5.1 billion active, it is designed for deep reasoning, accurate multi-step instruction following and complex agentic workflows. Its 131,072-token context window supports long documents, detailed analyses and extended conversations. It supports function calling, structured JSON output and streaming.


Technical specifications

  • PROVIDER LOCATION: Non-EU
  • PARAMETERS: 117B (5,1B active)
  • MAX CONTEXT: 131K token
  • LICENSE: Apache 2.0
  • FEATURES: Function calling, Reasoning
  • OUTPUT FORMATS: Raw_text, Json_object, Json_schema

Use cases

  • Multi-step reasoning on long documents
  • Autonomous agents for complex decisions
  • Legal and compliance analysis
  • Structured report generation

Implementation examples

Input
curl --request POST \
--url https://apiaruba.it/aiplatform/v1/chat/completions \
--header 'content-type: application/json' \
--header 'x-aruba-api-key: *****' \
--data '{
"model": "gpt-oss-120b",
"messages": [
{
"role": "system",
"content": "Sei un assistente conciso."
},
{
"role": "user",
"content": "Spiega cos'\''è un embedding."
}
],
"temperature": 0.7,
"max_tokens": 1000
}'
Output
{
"id": "chatcmpl-5c3c7148-63df-4255-a49d-b6811f1d326d",
"created": 1787560820,
"model": "gpt-oss-120b",
"object": "chat.completion",
"system_fingerprint": "vllm-0.24.0-0068fb53",
"choices": [
{
"finish_reason": "stop",
"index": 0,
"message": {
"content": "Un **embedding** è una rappresentazione numerica (solitamente un vettore di dimensione ridotta) di dati discreti o strutturati (parole, immagini, nodi di un grafo, ecc.) in uno spazio continuo. Lo scopo è catturare le relazioni semantiche o strutturali: elementi simili nel dominio originale hanno vettori vicini nella rappresentazione.\n\n- **Word embedding** (es. Word2Vec, GloVe): mappa parole in vettori dove la distanza riflette similarità di significato.\n- **Image embedding**: estratto da una rete neurale convoluzionale, consente confronti di immagini tramite distanza euclidea o coseno.\n- **Graph embedding**: mappa nodi o interi grafi in vettori per compiti come classificazione o clustering.\n\nGli embedding si apprendono tipicamente tramite reti neurali (auto‑encoder, modelli di language modeling, contrastive learning) o metodi di fattorizzazione matriciale, riducendo la dimensionalità e facilitando l’uso di algoritmi di machine learning tradizionali.",
"role": "assistant",
"reasoning_content": "The user asks in Italian: \"Spiega cos'è un embedding.\" They want explanation of what an embedding is. Concise. So respond in Italian, concise. Probably define embedding in machine learning, vector representation, mapping from discrete to continuous space, examples like word embeddings, image embeddings, etc. Keep concise.",
"provider_specific_fields": {
"reasoning": "The user asks in Italian: \"Spiega cos'è un embedding.\" They want explanation of what an embedding is. Concise. So respond in Italian, concise. Probably define embedding in machine learning, vector representation, mapping from discrete to continuous space, examples like word embeddings, image embeddings, etc. Keep concise.",
"refusal": null
}
},
"provider_specific_fields": {}
}
],
"usage": {
"completion_tokens": 307,
"prompt_tokens": 89,
"total_tokens": 396
}
}

Alternatives in the catalog


Cost

0,0319 € + VAT / 1M token (input)
0,1474 € + VAT / 1M token (output)

Resources