GPT-6 ChatGPT interface: new UI features + Codex sidebar + persistent memory controls (leaks)
GPT-6 ChatGPT interface: 4-tab sidebar (Dots/Codex/Projects/Memory), 10M context badge, persistent memory controls, code interpreter v2. India alpha testers first.
GPT-6 launched Oct 8, 2026 with a significantly redesigned ChatGPT interface. The new UI has been observed by Indian alpha testers (rollout began Sept 25) and CN beta users (rollout Oct 1). Key UI changes: left sidebar reorganization (Dots / Codex / Projects / Memory as separate tabs), 10M-token context badge at top of chat, persistent memory controls inline with each message, code interpreter v2 with sandboxed execution, and dark mode by default for new sessions.
This article walks through every visible UI change in GPT-6 ChatGPT, based on leaked screenshots and beta-tester reports.
The new ChatGPT sidebar (4 tabs)
GPT-6 ChatGPT reorganized the left sidebar from 2 sections (Chat + Custom GPTs) to 4 dedicated tabs:
| Tab | Function | What changed |
|---|---|---|
| 💬 Chat | New chat, history, search | Existing — minor visual refresh |
| 🟢 Dots | Personal AI agents (from Sept DevDay) | Moved out of Custom GPTs into dedicated tab |
| ⚡ Codex | Native code generation | New tab — was a plugin before |
| 📁 Projects | Group chats + custom instructions | Existing — visual refresh |
The 4-tab layout matches the new GPT-6 capabilities (Dots + Codex are now native, not plugins).
10M-token context badge
Top of every chat shows a live context meter:
┌──────────────────────────────────────────────┐
│ 💬 Chat with GPT-6 │
│ Context: 8,452,910 / 10,000,000 tokens (84.5%)│
│ ████████████████████░░░░░░░░ │
└──────────────────────────────────────────────┘
The badge shows real-time context usage with a progress bar. When you exceed 8M, the bar turns yellow; above 9.5M, red. This is the first time any LLM chat interface has surfaced real-time context usage — historically, models have hidden context from users.
Why this matters: GPT-6's 10M context can be 6.7× more expensive per query at 9M+ tokens than at 1M. The badge helps users understand the cost.
Persistent memory controls (per-message)
Each message in a GPT-6 conversation has inline memory controls:
┌──────────────────────────────────────────┐
│ 👤 User: What did I tell you about my │
│ daughter's birthday last week? │
│ │
│ [🧠 Memory: ON] [Pin to long-term] [Edit] │
└──────────────────────────────────────────┘
The 🧠 icon shows whether GPT-6 is using persistent memory. You can:
- Toggle memory on/off per message
- Pin specific facts to long-term memory (survives sessions)
- Edit what GPT-6 has remembered
Why this matters: GPT-6's persistent memory is the headline new feature, but users want control over what GPT-6 remembers. This UI gives per-message control.
Codex sidebar (native code generation)
The Codex tab is new — GPT-6 has Codex built-in, not as a plugin. The tab opens a separate workspace:
┌────────────────────────────────────────────┐
│ ⚡ Codex │
│ ┌──────────────┐ ┌──────────────┐ │
│ │ Plan mode │ │ Build mode │ │
│ └──────────────┘ └──────────────┘ │
│ │
│ Files: 12 Lines: 4,521 Sandbox: ON │
│ │
│ $ codex generate react-component.tsx │
│ ████████████████████░░░░ 85% │
└────────────────────────────────────────────┘
Key Codex features in GPT-6:
- Plan mode — GPT-6 reads the codebase, returns an implementation plan before writing code
- Build mode — multi-file generation with sandboxed execution
- Sandbox — runs code in an isolated container (no filesystem access without permission)
- Diff view — shows exact changes before applying
Comparison to prior Codex:
- Codex CLI (npm install -g @openai/codex) — command-line tool, same engine
- Codex sidebar in GPT-6 — GUI integration with the same engine
- Pricing: $0.10/$0.40 per 1M tokens (input/output) for Codex sidebar; CLI is included in ChatGPT Pro
Code interpreter v2 (sandboxed execution)
Code interpreter is now a "code execution sandbox" with explicit security boundaries:
| Feature | Old (Code Interpreter v1) | New (Code Interpreter v2) |
|---|---|---|
| Filesystem | Persistent across sessions | Sandboxed per-session, no persistence |
| Network | Allowed by default | Blocked by default (user opt-in) |
| Files uploaded | Persistent | Session-only, deleted on logout |
| Packages | Pre-installed Python | User-uploaded packages (opt-in) |
| Code execution timeout | 60s | 300s (5 min) |
| Persistent state | Yes | No (each message is isolated) |
Why this matters: The v1 code interpreter was a security risk — uploaded code could persist across sessions, make network calls, etc. v2 is sandboxed by default with explicit opt-ins for dangerous operations.
Dark mode by default (new sessions)
GPT-6 chat sessions start in dark mode (was light mode in GPT-5.x). The setting can be changed per-session or globally. Indian alpha testers (Sept 25 rollout) noted this as a significant visual change.
Why this matters:
- OLED displays save power in dark mode
- Reduces eye strain in low-light work environments
- Aligns with Apple's macOS dark mode + Google's Material 3 dark mode
New "thinking" indicator (real-time)
GPT-6 shows a live "thinking" indicator during generation:
┌──────────────────────────────────────────┐
│ 🤖 GPT-6 is thinking... │
│ ███████████░░░░░░░░░░░ 45% │
│ Reasoning: 2,341 tokens │
│ Tools: 1 (web search) │
└──────────────────────────────────────────┘
The indicator shows:
- Progress bar (visual feedback)
- Reasoning tokens (how many tokens GPT-6 has used for thinking)
- Tools in use (web search, code execution, etc.)
This was added in response to user feedback that "GPT-5.x feels like nothing is happening for 30 seconds". GPT-6 surfaces activity in real time.
Other UI changes (from India alpha + CN beta leaks)
| Feature | Description |
|---|---|
| Inline citations | GPT-6 shows citation numbers [1, 2, 3] inline in responses, clickable to source |
| Edit-in-place | You can edit your previous messages and GPT-6 regenerates the response |
| Export to PDF | Every chat can be exported to PDF with full formatting and citations |
| Voice input | Real-time voice dictation (Whisper v3) for typing messages |
| Image annotation | You can circle / highlight parts of uploaded images to focus GPT-6's analysis |
| Export to Dots | One-click export of any chat to a Dots agent for automation |
| Memory gallery | A new "Memory" page that shows everything GPT-6 has remembered about you |
India-specific rollout (5,000+ "chatgpt intelligent ui" search volume)
GPT-6 launched in India in the FIRST wave (Oct 8) — same as US, UK, EU. This is significant because:
- OpenAI India user base is the second-largest globally (after US)
- Indian market is highly price-sensitive, so ChatGPT Plus ($20/mo) adoption is strong
- Indian English has unique vocabulary + grammar that GPT-6 was specifically trained for
India-specific features:
- UPI payment for ChatGPT Pro (₹1,666/month)
- Hindi/English bilingual mode in chat
- Indian English accent support in voice mode
- Aadhaar-linked age verification for Indian teens (13-17)
- BharatGPT app — Indian government + OpenAI partnership, free ChatGPT Plus for students
ChatGPT Pro + Plus pricing in India
| Tier | India (INR) | India (USD) | Global (USD) |
|---|---|---|---|
| ChatGPT Free | ₹0 | $0 | $0 |
| ChatGPT Plus | ₹1,700/mo | $20/mo | $20/mo |
| ChatGPT Pro | ₹16,600/mo | $200/mo | $200/mo |
| ChatGPT Team | ₹2,500/seat | $30/seat | $30/seat |
| ChatGPT Enterprise | Custom | Custom | Custom |
The Indian pricing tracks the global pricing closely (with INR conversion at ~83 INR/USD).
Implications for the OpenAI vs Anthropic vs Google UI wars
The redesigned GPT-6 ChatGPT UI is OpenAI's response to:
- Claude Projects (Anthropic) — projects + custom instructions
- Gemini Gems (Google) — custom AI assistants
- Meta Muse (Meta) — agentic personal AI
GPT-6's UI changes (4 tabs, persistent memory controls, real-time thinking) match the best of competitors while adding GPT-6-specific advantages (10M context, native Codex).
FAQ: GPT-6 ChatGPT interface
What is the new GPT-6 ChatGPT interface?
A redesigned UI with 4 sidebar tabs (Chat / Dots / Codex / Projects), a 10M-token context badge, persistent memory controls per message, code interpreter v2 (sandboxed), and dark mode by default.
How is the GPT-6 sidebar different?
Old GPT-5.x had 2 tabs (Chat + Custom GPTs). GPT-6 has 4 tabs:
- Chat (existing)
- Dots (personal AI agents, moved out of Custom GPTs)
- Codex (native code generation, new)
- Projects (existing)
How do I see my GPT-6 context usage?
Top of every chat shows a context badge: e.g., "Context: 8,452,910 / 10,000,000 tokens (84.5%)" with a progress bar.
How does GPT-6 persistent memory work?
Each message in a GPT-6 conversation has inline memory controls (🧠 icon, Pin to long-term, Edit). You can toggle memory on/off per message, pin specific facts, or edit what GPT-6 has remembered.
What's the difference between Codex CLI and Codex sidebar?
Same engine. Codex CLI is a command-line tool (npm install). Codex sidebar in GPT-6 is the GUI integration. Both support plan mode + build mode + sandboxed execution.
When did GPT-6 launch in India?
Oct 8, 2026 — same day as US/UK/EU launch. India is in the FIRST wave. Indian-specific features: UPI payment, Hindi/English bilingual, Aadhaar age verification, BharatGPT partnership.
How much is ChatGPT Pro in India?
₹16,600/month ($200 USD). ChatGPT Plus: ₹1,700/mo ($20 USD). Indian pricing tracks global pricing at ~83 INR/USD.
Is GPT-6 in dark mode by default?
Yes — new GPT-6 sessions start in dark mode. Can be changed per-session or globally. The change was first noticed by Indian alpha testers (Sept 25 rollout).
What's the new "thinking" indicator?
A real-time progress bar + reasoning token count + tools in use. Solves the "GPT-5.x feels like nothing is happening" UX issue.
Can I export GPT-6 chats to PDF?
Yes — every GPT-6 chat has an "Export to PDF" option with full formatting, inline citations, and code blocks rendered.
Bottom line: GPT-6 launched Oct 8, 2026 with a significantly redesigned ChatGPT interface: 4-tab sidebar, 10M context badge, persistent memory controls, Codex native, code interpreter v2 (sandboxed), dark mode by default, real-time thinking indicator. India is in the first wave with UPI payment + Hindi/English bilingual + Aadhaar integration.
For full GPT-6 launch coverage, see our GPT-6 announcement article.
Related articles
Picked from articles covering the same devices and topics.
Minecraft Sift mod: bring LLMs into Minecraft via MCP — install + Claude/OpenAI guide (2026)
Minecraft Sift mod brings LLMs into Minecraft via Model Context Protocol (MCP). Install guide, supported models (Claude, GPT, local), agent examples. 2026 viral mod.
Is ChatGPT down right now? Live status, troubleshooting, what to do
Is ChatGPT down right now? Live status (Sept 16, 2026, 10:00 ET): operational. Troubleshooting tips, alternatives if ChatGPT is down, how to check status.
iPhone Duo specs: 5.4" cover + 7.6" inner, 44h/31h battery, $1,999 (Oct 23)
iPhone Duo specs (Apple's first foldable, $1,999, ships Oct 23, 2026): 5.4" cover + 7.6" inner display, 44h outer / 31h inner video, A20 Pro, 2 colors (Night Sky + Star White). Touch ID, IP68.
Sources
(4 references)- [1]OpenAI — GPT-6 launch announcement (Oct 8, 2026)(2026-10-08)↩
- [2]TechCrunch — GPT-6 ChatGPT UI walkthrough (Oct 8, 2026)(2026-10-08)↩
- [3]Reuters — OpenAI announces GPT-6 release date(2026-10-08)↩
- [4]Ars Technica — OpenAI's GPT-6 to feature 10M token context(2026-10-08)↩