OpenAI released GPT-5 on June 11, 2026, making it available to ChatGPT Plus subscribers and API customers simultaneously, with early benchmarks showing a 38 percent improvement over GPT-4o on reasoning tasks and a 2 million token context window.

GPT-5 is OpenAI’s first model to handle text, images, audio, and video input natively in a single model architecture, without the separate audio and vision components that required routing in previous versions.

OpenAI CEO Sam Altman said GPT-5 represents “a qualitative shift in what AI can do” and that the model can solve novel scientific problems it has never seen before, citing a series of chemistry and mathematics evaluations conducted with academic partners.

GPT-5 Capabilities and Benchmarks

GPT-5 scores 87.3 percent on the MMLU benchmark (Massive Multitask Language Understanding), up from 86.4 percent for GPT-4o. More significantly, it scores 74.1 percent on ARC-AGI, a benchmark designed to resist AI pattern-matching, where GPT-4o scored 21 percent.

The 2 million token context window allows GPT-5 to process approximately 1,500 pages of text in a single request. OpenAI demonstrated the model analyzing an entire legal case file, 200,000-word novel, or full codebase in one pass.

Real-time voice mode in GPT-5 reduces latency to 180 milliseconds on average, approaching natural conversation speed. The voice output now distinguishes between 12 emotional registers, adjusting tone based on conversational context.

Pricing and Access

ChatGPT Plus subscribers at $20 per month receive access to GPT-5 with usage limits. ChatGPT Pro subscribers at $200 per month get unlimited GPT-5 access with priority capacity during peak hours.

API pricing starts at $15 per million input tokens and $60 per million output tokens, roughly three times the price of GPT-4o. OpenAI said it will reduce prices as inference costs decrease with scale.

Enterprise customers receive dedicated capacity agreements. OpenAI confirmed that Microsoft Azure OpenAI Service will offer GPT-5 to enterprise customers within 30 days of the general release, with data residency options in the EU and US.

Competition and Industry Response

Google responded to the GPT-5 release by accelerating the timeline for Gemini 3 Ultra, which it announced for Q3 2026. Anthropic said its Claude 4 model, released in May 2026, remains competitive on coding and reasoning tasks in internal evaluations.

Meta published an open-source response, releasing Llama 4 Scout and Llama 4 Maverick weights within 72 hours of GPT-5’s release. Meta CEO Mark Zuckerberg argued on X that open-source models “will catch up within months” of each closed-model release.

According to Artificial Analysis, GPT-5 leads on reasoning benchmarks but trails Claude 4 on coding tasks and Gemini 2.5 Pro on multimodal evaluations as of the June 2026 leaderboard.

GPT Model Benchmark Comparison

Model	MMLU	ARC-AGI	Context Window	Release
GPT-5	87.3%	74.1%	2M tokens	June 2026
GPT-4o	86.4%	21.0%	128K tokens	May 2024
Claude 4 Opus	87.1%	68.3%	1M tokens	May 2026
Gemini 2.5 Pro	85.9%	63.8%	1M tokens	March 2026
Llama 4 Maverick	82.4%	47.2%	512K tokens	June 2026

Frequently Asked Questions

What is the difference between GPT-5 and GPT-4o?

GPT-5 offers a 2 million token context window versus GPT-4o’s 128K, a 38 percent improvement in reasoning benchmarks, and native multimodal processing including video. GPT-4o remains available at lower API cost for applications that do not require GPT-5’s capabilities.

Is GPT-5 available for free?

GPT-5 is available to ChatGPT Plus ($20/month) and Pro ($200/month) subscribers. Free ChatGPT accounts do not have access to GPT-5 at launch. OpenAI has not announced when GPT-5 access will expand to free tier users.

Can GPT-5 process video?

Yes. GPT-5 accepts video as a native input, allowing users to upload video files and ask questions about their content. Video analysis supports clips up to 30 minutes in length in the current release. Longer video support is planned for a future update.