Curiosity · AI Model

GPT-5

GPT-5 is OpenAI's 2026 flagship large language model — a unified multimodal system with native audio, vision, and code, a switchable reasoning mode, and the deepest developer ecosystem of any model in the market. It is the default general-purpose choice for most teams evaluating AI infrastructure.

Model specs

Vendor: OpenAI
Family: GPT-5
Released: 2025-08
Context window: 400,000 tokens
Modalities: text, vision, audio, code
Input price: $10/M tok
Output price: $40/M tok
Pricing as of: 2026-04-20

Strengths

Broadest ecosystem — Responses API, Realtime API, Assistants, Batch, Structured Outputs
Best-in-class audio (native speech in/out), competitive vision
Strong instruction following with JSON-mode + structured outputs

Limitations

Slightly behind Claude Opus 4.7 on long-horizon coding agents
Thinking mode adds noticeable latency on hard problems
Context window smaller than Claude Opus 4.7's 1M

Use cases

General-purpose chat and copilot assistants
Voice / real-time audio agents via Realtime API
Multimodal understanding — image, PDF, screen capture
Enterprise deployments that standardise on OpenAI / Azure OpenAI

Benchmarks

Benchmark	Score	As of
MMLU-Pro	≈87%	2026-04
SWE-bench Verified	≈70%	2026-04
AIME 2025	≈94%	2026-04

Frequently asked questions

What is GPT-5?

GPT-5 is OpenAI's 2026 flagship large language model — a unified multimodal system with text, vision, audio, and code, a switchable reasoning mode, and the deepest developer ecosystem of any frontier model.

How does GPT-5 compare to Claude Opus 4.7?

GPT-5 is stronger on multimodal (particularly audio) and has a broader API surface. Claude Opus 4.7 leads on long-context and long-horizon coding agents and has a larger 1M-token context window.

What is GPT-5's context window?

As of April 2026, GPT-5 offers a 400,000-token context window across the Responses API. Batch and long-context modes are available for specific use cases.

Sources

OpenAI — Models — accessed 2026-04-20
OpenAI — Pricing — accessed 2026-04-20