OpenAI GPT-5 - AI Language Models Tool

Overview

OpenAI GPT-5 is OpenAI’s flagship language model family (released August 7, 2025) optimized for advanced reasoning, code generation, instruction following, and agentic tool use. The GPT-5 family is offered in multiple sizes—gpt-5, gpt-5-mini, and gpt-5-nano—and includes configurable reasoning effort and verbosity controls so developers can trade off thinking time, quality, and latency for their applications. GPT-5 also exposes new tool primitives (including a "custom" freeform tool type), built-in tools (web search, file search, image generation), and explicit support for long-context retrieval and high-quality agentic workflows. ([openai.com](https://openai.com/index/introducing-gpt-5/)) For developers, OpenAI provides GPT-5 via the Responses and Chat Completions APIs and a web playground; the API adds explicit parameters to control model behavior (reasoning_effort: minimal/low/medium/high and verbosity: low/medium/high), prompt caching and batch APIs for cost reductions, streaming outputs, and examples for integrating GPT-5 into coding, frontend, and agentic pipelines. OpenAI documents SOTA benchmark gains in coding, math, multimodal understanding, and tool-calling, and publishes tokenized pricing tiers for the gpt-5 family on the platform pricing page. Use cases include end-to-end agentic automation, frontend prototyping from a single prompt, and high-precision code editing and debugging. ([openai.com](https://openai.com/index/introducing-gpt-5-for-developers))

Key Features

• Configurable reasoning depth (minimal/low/medium/high) to trade speed for accuracy. ([openai.com](https://openai.com/index/introducing-gpt-5-for-developers))
• Verbosity control (low/medium/high) to steer answer length and token usage. ([openai.com](https://openai.com/index/introducing-gpt-5-for-developers))
• Three model sizes: gpt-5, gpt-5-mini, gpt-5-nano for cost/latency tradeoffs. ([openai.com](https://openai.com/index/introducing-gpt-5-for-developers))
• Custom freeform tools allow raw text payloads to external tools (code sandboxes, SQL, shells). ([cookbook.openai.com](https://cookbook.openai.com/examples/gpt-5/gpt-5_new_params_and_tools?utm_source=openai))
• Built-in tools (web search, file search, image generation) and parallel tool-calling for agentic tasks. ([openai.com](https://openai.com/index/introducing-gpt-5-for-developers))
• Long-context retrieval and evaluations up to 256k token ranges for multi-document tasks. ([openai.com](https://openai.com/index/introducing-gpt-5-for-developers))
• Prompt caching and Batch API to reduce latency and input costs for repeated prompts. ([platform.openai.com](https://platform.openai.com/docs/guides/prompt-caching/prompt-caching%20.pls?utm_source=openai))

Example Usage

Example (python):

from openai import OpenAI
import os

# set OPENAI_API_KEY in your environment
client = OpenAI()

# Simple Responses API call using GPT-5 with reasoning and verbosity control
resp = client.responses.create(
    model="gpt-5",
    input=[
        {"role": "user", "content": "Audit this Python function and suggest bug fixes and tests."}
    ],
    # control thinking depth and output length
    reasoning={"effort": "high"},
    text={"verbosity": "medium"},
    stream=False,
)

print(resp.output_text)

# For streaming (long outputs) set stream=True and iterate events.
# This example follows OpenAI Responses API patterns and GPT-5 parameters documented by OpenAI. ([platform.openai.com](https://platform.openai.com/docs/guides/Reasoning?api-mode=response&utm_source=openai))

Pricing

OpenAI publishes per-1M-token pricing for the GPT-5 family on the platform pricing page: gpt-5 input $1.25 / 1M tokens, output $10.00 / 1M tokens; gpt-5-mini input $0.25 / 1M, output $2.00 / 1M; gpt-5-nano input $0.05 / 1M, output $0.40 / 1M. Pricing table and cached-input/Batch tiers are detailed on OpenAI’s pricing docs. ([platform.openai.com](https://platform.openai.com/docs/pricing/?utm_source=openai))

Benchmarks

SWE-bench Verified (real-world coding): 74.9% (Source: https://openai.com/index/introducing-gpt-5-for-developers/)

Aider Polyglot (code editing): 88% (Source: https://openai.com/index/introducing-gpt-5-for-developers/)

AIME (math, without tools): 94.6% (Source: https://openai.com/index/introducing-gpt-5/)

GPQA (GPT-5 pro, challenging science questions): 88.4% (Source: https://openai.com/index/introducing-gpt-5/)

τ2-bench telecom (tool-calling): 96.7% (Source: https://openai.com/index/introducing-gpt-5-for-developers/)

Key Information

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