ChatGPT New Features 2026 — GPT-5.5, Every Update (June 2026)

Who gets it: Paid plans (Go, Plus, Pro) — Free users stay on automatic model selection. Rolling out in India alongside global users.

On 10 June 2026, OpenAI replaced the confusing pile of “Thinking” variants with six plain-language options: Instant, Medium, High, Extra High, Pro Standard, and Pro Extended. The old names map directly — Thinking Standard became Medium, Thinking Extended became High, and Thinking Heavy became Extra High. Thinking-Light was removed entirely; OpenAI said fewer than 1% of paid subscribers ever used it.

The picker now reads as a speed-versus-depth slider rather than a model catalogue. You no longer need to know what “GPT-5.5 Thinking Heavy” means — you just pick how hard you want it to think.

Who gets it: All personal accounts — Free, Go, Plus and Pro — plus self-serve ChatGPT Business. Available in India.

Announced on 4 June 2026, Lockdown Mode is an optional setting for people working with sensitive data. When enabled, it switches ChatGPT into a conservative configuration: live web access, image support in responses, Deep Research, Agent Mode, connectors, and file downloads are limited or turned off. The goal is to shut down the routes a prompt injection attack could use to exfiltrate your data — for example, malicious instructions hidden inside a webpage or uploaded file.

Alongside it, OpenAI introduced Elevated Risk labels — warnings shown when a feature you are about to use carries higher data-exposure risk. OpenAI is candid that Lockdown Mode is not a complete shield: injected instructions in cached content or uploaded files can still influence a response’s behaviour or accuracy.

Who gets it: Plus and Pro in the US from 4 June 2026; rolling out to more countries (including India) and to Free and Go users over the following weeks.

In a blog post titled “Dreaming: Better memory for a more helpful ChatGPT” (4 June 2026), OpenAI announced a rebuilt memory system that periodically re-synthesises what it knows about you instead of hoarding stale notes. Memory is now time-aware — “You are going to Singapore in July” becomes “You went to Singapore in July 2026” once the trip has passed. Plus and Pro users get double the memory capacity.

OpenAI’s own evaluations show factual recall improving from 67.9% to 82.8%, preference adherence from 55.3% to 71.3%, and accuracy-over-time from 52.2% to 75.1%. Just as important: a roughly 5x cut in compute cost means the new system is coming to Free and Go users too — the first time full memory synthesis reaches the free tier. You also get a readable memory summary page and controls to edit what ChatGPT remembers.

Who gets it: Everyone — including Free and Go users worldwide, India included.

Twelve days after GPT-5.5’s paid launch, OpenAI released GPT-5.5 Instant on 5 May 2026 and made it the default model for all ChatGPT users, replacing GPT-5.3 Instant. This is the speed-optimised variant: it keeps GPT-5.5’s quality gains while answering at the low latency free users expect.

The practical effect: for the first time since the GPT-5 launch cycle, a Free user in India and a Pro user in San Francisco start from the same base model. The differences between plans are now about usage limits, reasoning depth (the Thinking/High tiers), and extras — not about being stuck on an older brain.

Who gets it: GPT-5.5 — Plus, Pro, Business, Enterprise from 23 April 2026. GPT-5.5 Pro — Pro, Business, Enterprise. API access followed on 24 April 2026.

GPT-5.5 launched on 23 April 2026 as OpenAI’s new flagship, with the company calling it its strongest agentic coding model to date. The headline claims: noticeably reduced hallucination in high-stakes domains — law, medicine, finance — while keeping the latency of its predecessor. Notably, OpenAI withheld API access for one day at launch, saying API deployments required “different safeguards”; both GPT-5.5 and GPT-5.5 Pro reached the API on 24 April.

The release pace deserves comment: GPT-5.5 arrived only weeks after GPT-5.4, and barely five months after GPT-5.1. OpenAI is now shipping point releases the way phone makers ship camera updates — which is exactly why this page exists.

Who gets it: Rolling out to all ChatGPT users (Free with limits, higher limits on Go/Plus/Pro) from 22 April 2026; API as gpt-image-2.

Announced 21 April 2026, Images 2.0 is the first OpenAI image model that reasons before it draws — it can pull live information from the web, plan the image’s structure, and then render. It supports 2K resolution, aspect ratios from 3:1 to 1:3, and can generate up to eight coherent images from one prompt, keeping characters and objects consistent across the batch.

That last capability — character consistency across a batch — is what creators had been faking with elaborate prompt chains for two years. For Indian YouTubers, Instagram creators, and small D2C brands making product shots, it is the single most practical creative upgrade of 2026.

Who is affected: Anyone who still manually selected legacy models — all plans, all regions.

On 13 February 2026, OpenAI retired GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini from ChatGPT, alongside GPT-5 Instant and GPT-5 Thinking. GPT-4o’s removal closed the chapter that began ChatGPT’s voice-and-vision era in May 2024 — and the model whose “personality” users famously protested to keep during the GPT-5 launch. The models live on in the API, but not in the ChatGPT app.

The transition had buffers: Enterprise workspaces kept GPT-5 models until 19 February 2026, GPT-5 Pro stayed for all paid users until 19 February, and Business/Enterprise/Edu customers could use GPT-4o inside Custom GPTs until 3 April 2026. GPT-5.1 followed on 11 March 2026.

Two more retirements are scheduled: GPT-4.5 retires on 27 June 2026 (30-day sunset) and o3 — the last o-series reasoning model — retires on 26 August 2026 after a 90-day sunset period.

Who got it: All ChatGPT users; API as gpt-image-1.5 with 20% cheaper image input/output.

Launched 16 December 2025: roughly 4x faster generation, dramatically better text rendering inside images, and — the big one — edits that change only what you asked for instead of re-rendering the whole image. Lighting, composition and faces stay consistent across edits. A dedicated Images tab arrived with templates and trend-based filters, letting you generate without writing a prompt at all.

Who got it: All users, web and mobile, as the new default.

From 25 November 2025, talking to ChatGPT no longer threw you into the separate blue-orb screen. Voice now runs inside the conversation: you speak, watch the answer appear as text in real time, scroll earlier messages, and see images or maps mid-conversation. Prefer the old orb? Settings → Voice Mode → Separate mode brings it back.

Who got it: Paid plans from 13 November 2025, then Free and logged-out users.

Announced 12 November 2025, GPT-5.1 brought “warmer” responses and the tone preset system that survives today: Default, Friendly, Efficient, Professional, Candid, and Quirky, with experimental sliders for warmth, conciseness, and emoji frequency. The model itself was retired on 11 March 2026, but the personalization controls it introduced carried forward into GPT-5.5.

Who got it: India-first launch (August 2025 at ₹399/month with UPI), then 170+ more countries; ~$8/month in the US.

ChatGPT Go launched in India in August 2025 — OpenAI’s first India-first product decision — offering 10x more messages, file uploads, and image generation than Free, plus longer memory and context, Tasks, Projects, and custom GPTs, payable by UPI. Then on 4 November 2025, OpenAI made Go free for 12 months for users in India as a limited-time promotion, redeemable via the web app and Google Play (App Store a week later), with a payment method required and a refundable ₹2 verification charge.

Go has since become OpenAI’s fastest-growing plan globally. If you claimed the free year in November 2025, note that auto-billing at ₹399/month begins when your 12 months end — around November 2026. Set a reminder.

Who got it: Logged-in users on web first (4 September 2025), iOS and Android later in 2025.

Hover over any message, choose Branch in new chat, and ChatGPT forks the conversation from that point — original thread untouched, full context carried over. It is version control for conversations: test three different directions for the same brief without polluting your main thread.

Who gets it: Go, Plus, and Pro plans; works on Web, iOS, Android, and macOS (Windows app support on the roadmap, per OpenAI). Tasks management page is web-only.

Tasks lets ChatGPT run prompts on a schedule — one-off or recurring — and notify you by push or email when done, whether or not you are online. A daily 7 AM news brief, a weekly market summary, English practice every evening: set it once at Tasks in ChatGPT (official docs) and forget it. Through 2025–2026 it expanded across platforms and, crucially for India, is included in the ₹399 Go plan rather than gated to Plus.