GPT-5 Launch Brings Major Upgrades to Free Users

Key Points

• OpenAI rolls out GPT-5, its most advanced AI model to date
• GPT-5 blends real-time speed with deep reasoning capabilities
• ChatGPT free users now get access to GPT-5 by default
• GPT-5 beats rivals in coding, creativity, and health accuracy

OpenAI has officially launched GPT-5,  its most advanced AI model yet,  now available to all ChatGPT free users starting August 7. The release marks a major shift in the AI landscape, blending the speed of previous GPT models with the advanced reasoning power of OpenAI’s o-series.

Source: OpenAI

The model is described as OpenAI’s first “unified model,” meaning it can decide in real-time whether to respond quickly or take extra time to “think” through more complex tasks. It essentially combines the best of both speed and intelligence,  no extra settings or manual model switching required.

“It’s a significant step toward AGI,” said OpenAI CEO Sam Altman. “Having something like GPT-5 would have been unimaginable just a few years ago.”

GPT-5 is here.

Rolling out to everyone starting today.https://t.co/rOcZ8J2btI pic.twitter.com/dk6zLTe04s

— OpenAI (@OpenAI) August 7, 2025

OpenAI’s long-standing mission to make AI accessible is now clearer than ever. GPT-5 is being rolled out to all free users, giving millions access to powerful reasoning tools that were once behind a paywall. As ChatGPT now reaches over 700 million weekly users, this move is expected to dramatically shift how people use AI daily.

GPT-5 is also a key part of OpenAI’s plan to move ChatGPT from just a chatbot to a real AI agent,  one that can complete tasks, manage schedules, build apps, and generate research on its own.

GPT-5 is here — and we’re already putting it to work. ✨

We’re excited to be one of the first to integrate @OpenAI’s GPT-5 into our AI stack, bringing next-gen power to Zoom AI Companion.

With our federated AI approach (Zoom’s own models + third-party LLMs like GPT-5), AI… pic.twitter.com/jl2hgvOlUk

— Zoom (@Zoom) August 7, 2025

GPT-5 Takes the Lead in Key Benchmarks

With GPT-5, OpenAI has delivered clear gains in performance across some of the toughest AI tests. It’s designed to outperform previous models in real-world tasks like coding, healthcare responses, and creative writing — areas where AI is already seeing mass adoption.

🚨 BREAKING: GPT-5 is HERE — and it’s unreal.

– 4 new variants (GPT-5, Mini, Nano, Chat)
– PhD-level reasoning 🤯
– 75% on SWE Bench
– 90% cost drop
– Code, chat, build apps — FASTER than ever

The AI game just changed.#GPT5 #AI #OpenAI #AGI pic.twitter.com/FxmWjbklHp

— shubham vedi (@shub_vedi) August 7, 2025

Coding Excellence:
GPT-5 scored 74.9% on SWE-bench Verified, a coding benchmark using real GitHub tasks. It slightly beats Claude Opus 4.1 (74.5%) and outpaces Google’s Gemini 2.5 Pro (59.6%). Developers are now seeing GPT-5 spin up full software apps using what’s being called “vibe coding.”

Health and Safety:
In the HealthBench Hard Hallucinations test, GPT-5 with extended reasoning hallucinates only 1.6% of the time,  a major improvement over GPT-4o (12.9%) and o3 (15.8%). This matters as more users turn to AI for health-related information, even though chatbots aren’t licensed professionals.

Creative Edge:
OpenAI says GPT-5 shows better “taste” and natural flow in tasks like storytelling, branding, and design. This creative improvement pairs well with the growing number of ChatGPT conversations users are having that require more human-like responses.

Reduced Hallucinations:
Across general use, GPT-5 (with reasoning enabled) has a hallucination rate of just 4.8%, way down from GPT-4’s 20.6%. This builds trust and enhances user experience across both casual and professional use.

Science & Reasoning:
On the GPQA Diamond test (PhD-level science), GPT-5 Pro scored 89.4%, outperforming competitors like Claude Opus 4.1 (80.9%) and Grok 4 Heavy (88.9%).

Source: OpenAI

However, GPT-5’s performance isn’t perfect. On the Tau-bench,  a benchmark for agent-like behavior,  GPT-5 showed mixed results. It scored 63.5% on airline website navigation (slightly below o3 at 64.8%) and 81.1% in retail scenarios (behind Claude’s 82.4%).

Still, OpenAI claims GPT-5 is safer and more reliable. It identifies bad actors more accurately and reduces the number of false flags that block helpful user requests. The model’s honesty and transparency have improved, says Alex Beutel, OpenAI’s head of safety.

I have had early access to @OpenAI‘s GPT-5 for the last two weeks and it is the smartest model available as of now.

Just as one example, here it created an anamorphic text illusion in minecraft that spells “SPECIAL” from one angle and “GPTFIVE” from another pic.twitter.com/WUu75rEUrL

— adi (@adonis_singh) August 7, 2025

Features for Developers and Power Users

ChatGPT isn’t just smarter,  it’s now more fun and flexible, thanks to user experience upgrades and developer tools that ship with GPT-5.

New Personalities:
Users can choose between four built-in modes,  Cynic, Robot, Listener, and Nerd,  which alter GPT-5’s tone and style of responses without needing extra prompts.

Flexible Plans:

• Free Users: Get default access to GPT-5

• Plus Plan ($20/month): Higher usage limits

• Pro Plan ($200/month): Unlimited use + access to GPT-5 Pro

• Enterprise Users: Teams, Edu, and Enterprise plans get GPT-5 as the default next week

Developer Tools:
For developers, GPT-5 arrives via the OpenAI API in three formats:

• gpt-5 (full power)

• gpt-5-mini (faster, lighter)

• gpt-5-nano (for lightweight tasks)

Pricing starts at $1.25 per million input tokens and $10 per million output tokens. Developers can now also control verbosity, deciding how long or short GPT-5’s replies should be, which is especially useful for apps and integrations.

GPT-5 launches shortly after OpenAI released GPT-oss, its open-weight reasoning model that can run independently and offers near GPT-4 performance at a lower cost. However, GPT-5 remains the go-to for cutting-edge capabilities, especially in complex tasks and creative work.

Meanwhile, AI education tools are catching up. Platforms like Google’s guided learning for AI show how big tech is racing to keep users informed and safe in the new AI era.