PALO ALTO, March 18 — OpenAI released GPT-5 to ChatGPT Plus subscribers Wednesday, the company announced in a blog post, marking its first major model upgrade since the launch of GPT-4o in May 2024. The new model includes native multimodal reasoning across text, image, and audio, longer tool-use chains, and a context window the company said extends to 1 million tokens for paid API customers.
The release follows roughly twenty months of public anticipation and a series of incremental upgrades to GPT-4o that OpenAI had positioned as bridging models. GPT-5 is being made available immediately to Plus, Pro, Team, and Enterprise subscribers in the ChatGPT product, with API access rolling out to developers over the next two weeks, the company said.
“GPT-5 is a step change in how the model reasons across modalities and across long tool-use chains,” an OpenAI spokesperson said in response to questions from Consumer Tech Wire. The spokesperson declined to specify total parameter count or training compute, citing competitive considerations, but said the model is “substantially more efficient at inference” than GPT-4-class systems for comparable task quality.
What changed
In its release notes, OpenAI highlighted three areas of improvement. First, the model performs natively in mixed-modality contexts — meaning a single prompt can interleave text, images, and audio without separate routing, a workflow that previously required GPT-4o plus the company’s voice and vision modules. Second, agentic tool use has been extended: the company said GPT-5 can plan and execute tool-calling sequences of “20-plus distinct steps” without losing the original task framing. Third, the context window for API customers extends to 1 million tokens, with a 400,000-token window in the ChatGPT product itself.
OpenAI also said the model’s hallucination rate on its internal factual-recall evaluations is “approximately 45 percent lower” than GPT-4o, though the company did not release the underlying benchmark numbers.
Pricing and availability
ChatGPT Plus pricing remains $20 per month and Pro remains $200 per month. API pricing for GPT-5 was set at $5.00 per million input tokens and $15.00 per million output tokens, roughly in line with the GPT-4o pricing OpenAI introduced in 2024. A smaller “GPT-5 mini” model, intended for high-volume API workloads, was priced at $0.50 per million input tokens.
Free-tier ChatGPT users will receive limited GPT-5 access on a daily message cap, the company said, with overflow traffic falling back to GPT-4o.
Reaction
Anthropic, OpenAI’s most direct competitor, declined to comment on the GPT-5 release. The company shipped Claude Opus 4 in January and is widely expected to release a successor model in the second half of 2026.
“The interesting question for developers is not which model is marginally better on a given evaluation,” said Pradeep Lakshminarayan, a partner at the venture firm Conviction who tracks the foundation-model market. “It is whether the unit economics of inference have moved enough that the agentic workloads people have been prototyping for two years actually become deployable. GPT-5 mini at fifty cents per million input tokens is the more important number on this release page.”
OpenAI said it will host a developer event in San Francisco on April 14 to demonstrate GPT-5 in agentic and multimodal workflows.
Asari Whitfield-Asari covers AI tools and developer infrastructure for Consumer Tech Wire.