OpenAI GPT-5 - AI Language Models Tool

Overview

OpenAI GPT-5 is OpenAI’s next-generation family of large language models built for advanced multi-step reasoning, code generation, instruction following, and reliable tool use. The GPT‑5 family is offered in multiple sizes (gpt-5 / gpt-5-mini / gpt-5-nano) so teams can trade off cost, latency, and capability: the full gpt-5 is aimed at complex reasoning and coding tasks while mini/nano serve chat, classification, and low-latency needs. ([replicate.com](https://replicate.com/openai/gpt-5?utm_source=openai)) For developers, GPT‑5 exposes new steering knobs (reasoning_effort and verbosity) and improved tool integration so the model can chain dozens of tool calls reliably and handle tool errors more robustly. GPT‑5 is the default model in the ChatGPT experience and is available through the Responses and Chat Completions APIs; OpenAI also publishes a pro variant optimized for the highest-quality, long-horizon reasoning. The release emphasizes safety (extensive red‑teaming and biological-risk mitigations) and includes presets and API controls to reduce hallucinations and improve instruction following. ([openai.com](https://openai.com/index/introducing-gpt-5-for-developers?utm_source=openai))

Key Features

• Multi-size family: gpt-5, gpt-5-mini, gpt-5-nano for accuracy vs cost tradeoffs.
• Reasoning control: reasoning_effort (minimal/low/medium/high) to tune thought time.
• Verbosity steering: verbosity parameter to request concise or comprehensive outputs.
• Robust tool use: chains dozens of tool calls with improved error handling and routing.
• Safety-first: extensive red‑teaming and preset safeguards for high‑risk domains.

Example Usage

Example (python):

# Example: call GPT-5 with reasoning and verbosity controls (OpenAI Python client)
# See OpenAI docs for Responses API and parameters (reasoning_effort, verbosity). ([openai.com](https://openai.com/index/introducing-gpt-5-for-developers?utm_source=openai))
from openai import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-5",
    input="Write a tested Python function that converts Markdown to HTML and explain edge cases.",
    reasoning_effort="high",   # values: minimal, low, medium, high
    verbosity="high"           # values: low, medium, high
)

# Print textual output (response structure may vary by SDK version)
print(response.output[0].content[0].text)

Pricing

OpenAI lists per‑token pricing for the GPT‑5 family on its pricing page; recent published rates (example variants) include GPT‑5.2 at $1.750 per 1M input tokens and $14.000 per 1M output tokens, GPT‑5.2 Pro at higher professional rates, and GPT‑5 mini at $0.250 per 1M input and $2.000 per 1M output. Pricing, cached-input rates, and batch/enterprise terms appear on OpenAI’s API pricing page; confirm current rates on OpenAI’s pricing documentation. ([openai.com](https://openai.com/api/pricing/?utm_source=openai))

Benchmarks

External expert preference (GPT‑5 pro vs GPT‑5 thinking): 67.8% of evaluated prompts preferred GPT‑5 pro (Source: https://openai.com/index/introducing-gpt-5/)

Major error reduction (GPT‑5 pro vs GPT‑5 thinking): 22% fewer major errors in internal evaluations (Source: https://openai.com/index/introducing-gpt-5/)

Safety red‑teaming effort: 5,000 hours of red-team testing for biological-risk preparedness (Source: https://openai.com/index/introducing-gpt-5)

Key Information

• Category: Language Models
• Type: AI Language Models Tool