ChatGPT 4.5 (Orion): What It Was, Why It's Retired & What to Use Now (2026)

ChatGPT 4.5 (codenamed "Orion") was OpenAI's largest non-reasoning model, released as a research preview on February 27, 2025, and retired from ChatGPT Plus, Team, and the API on August 7, 2025 alongside the GPT-5 launch. GPT-4.5 was positioned as OpenAI's most knowledgeable writing and conversation model, with a notably lower hallucination rate than GPT-4o but no chain-of-thought reasoning. In 2026, GPT-4.5 is no longer available to most users. ChatGPT now runs on GPT-5.3 Instant (the default for every tier) and GPT-5.4 Thinking / Pro (for deeper reasoning), which have replaced GPT-4.5's role for writing, research, and knowledge tasks. AI Toolbox (formerly ChatGPT Toolbox), a Chrome extension with 20,000+ active users and a 4.5/5 Chrome Web Store rating, works with every current ChatGPT model and adds folders, full-text search, bulk export, prompt library, and prompt chaining to organize conversations across models.

If you landed here searching "ChatGPT 4.5" in 2026, you're most likely looking for a model that no longer exists in the ChatGPT app for Plus and Team users. This guide explains what GPT-4.5 was, what it did well, why OpenAI retired it, and which current model (GPT-5.3 Instant, GPT-5.4 Thinking, or GPT-5.4 Pro) is the right replacement for your task. If you want the full current lineup, read our complete ChatGPT models guide (2026).

What Was ChatGPT 4.5 (Orion)?

GPT-4.5 was OpenAI's largest non-reasoning model, designed for natural writing, broad world knowledge, and emotional-intelligence tasks where chain-of-thought reasoning was not required. It was released on February 27, 2025 as a research preview, and was the last major GPT-4 series model before OpenAI shifted to the GPT-5 line.

Unlike o-series models (o1, o3), GPT-4.5 did not use internal chain-of-thought reasoning. It responded directly from your prompt, which made it faster than reasoning models but less accurate on multi-step logic tasks. OpenAI described it as offering "broader world knowledge" and "improved ability to follow user intent" compared to GPT-4 and GPT-4o. On OpenAI's SimpleQA benchmark, GPT-4.5 hallucinated 37.1% of the time versus 61.8% for GPT-4o (OpenAI System Card, 2025) - a meaningful reduction but still far from perfect.

GPT-4.5 was best understood as a "writer and thinker" model rather than a "problem solver." It excelled at long-form prose, nuanced conversation, brainstorming, and summarization. It was weaker than o-series models on coding, math, and scientific reasoning. Many writers and researchers preferred it over GPT-4o for drafting articles, grant proposals, and long-form analysis.

When Was ChatGPT 4.5 Retired?

OpenAI retired GPT-4.5 from the API and ChatGPT Plus/Team tiers on August 7, 2025 when GPT-5 launched. Pro users retained access via a "Legacy Models" tab for a limited window. As of April 2026, GPT-4.5 is not available in ChatGPT for the vast majority of users.

The retirement timeline:

• February 27, 2025: OpenAI releases GPT-4.5 as a research preview for Pro subscribers and API users
• March 2025: GPT-4.5 rolls out to ChatGPT Plus, Team, and Enterprise
• July 14, 2025: GPT-4.5 removed from OpenAI API and Developer Playground
• August 7, 2025: GPT-5 launches; GPT-4.5 removed from ChatGPT Plus and Team. Pro users access it via "Legacy Models"
• February 13, 2026: OpenAI retires GPT-4o, GPT-4.1, GPT-4.1 mini, o4-mini, and GPT-5 (Instant and Thinking) from ChatGPT. GPT-5.3 / GPT-5.4 take over
• April 2026: GPT-4.5 is no longer a practical option for new work in ChatGPT

The core reason GPT-4.5 was retired: OpenAI's GPT-5 line absorbed GPT-4.5's strengths (writing quality, low hallucination rate, natural tone) while adding reasoning capability and a unified routing system that picks the right model per query. Maintaining GPT-4.5 as a separate option no longer made sense once GPT-5 surpassed it across most benchmarks.

What Replaced ChatGPT 4.5? (Current 2026 Models)

ChatGPT 4.5's role has been split between GPT-5.3 Instant (everyday writing and knowledge), GPT-5.4 Thinking (reasoning-heavy tasks), and GPT-5.4 Pro (hardest tasks and long workflows). Here's how the current lineup maps to what GPT-4.5 used to do:

For a full side-by-side comparison of every current ChatGPT model, read our ChatGPT models explained guide (2026).

How GPT-5.3 Instant and GPT-5.4 Compare to GPT-4.5

GPT-5.3 Instant and GPT-5.4 Thinking outperform GPT-4.5 on almost every dimension: writing quality, knowledge accuracy, reasoning, tool use, and availability. The one thing GPT-4.5 had that current models lack is its specific "large non-reasoning" positioning - but GPT-5.3's auto-switching router and GPT-5.4 Thinking's chain-of-thought reasoning more than make up for it.

The biggest shift: GPT-5.3 Instant is available to free users, bringing high-quality writing and knowledge to everyone. GPT-4.5 was never free - it required ChatGPT Plus at minimum.

Can I Still Access ChatGPT 4.5 in 2026?

No, ChatGPT 4.5 is not available to ChatGPT Plus, Team, Business, or Free users in 2026. ChatGPT Pro subscribers may find it under the "Legacy Models" tab, though OpenAI's legacy window for older models is limited. As of April 2026, OpenAI's active lineup only includes the GPT-5.3 and GPT-5.4 families.

To check whether any legacy model is still accessible in your ChatGPT Pro account:

1. Sign in to chatgpt.com with a ChatGPT Pro account
2. Start a new chat
3. Click the model picker at the top of the chat window
4. Click Configure in the picker
5. Toggle "Access legacy models" if available
6. Look for GPT-4.5 in the legacy list (may or may not be present)

For API developers, GPT-4.5 was removed from the OpenAI API on July 14, 2025 and is no longer available for new projects. If your codebase depends on GPT-4.5 via the API, migrate to gpt-5.4 or gpt-5.3-instant, which cover GPT-4.5's use cases and outperform it on OpenAI's benchmarks.

Should You Still Search for "GPT-4.5" Content?

No. Any GPT-4.5 tips, prompt templates, or pricing benchmarks published before August 2025 are outdated for current ChatGPT use. Here's what's still useful versus what to ignore:

• Still useful: Prompt engineering principles, long-form writing techniques, workflow advice - these transfer to GPT-5.3 and GPT-5.4
• Outdated - ignore: GPT-4.5 pricing, API rate limits, benchmark numbers, model picker paths, weekly message quotas
• Outdated - replace: Model recommendations (use GPT-5.3 Instant, GPT-5.4 Thinking, or GPT-5.4 Pro instead)

The fastest way to get accurate 2026 ChatGPT info is to check the official OpenAI help center article on GPT-5.3 and GPT-5.4.

Running writing and research projects across multiple ChatGPT models?

ChatGPT Toolbox adds folders, full-text search, and bulk export to ChatGPT so you can organize GPT-5.3 Instant drafts separately from GPT-5.4 Thinking research threads. Trusted by 20,000+ users with a 4.5/5 Chrome Web Store rating. Install free.

GPT-4.5's Lasting Impact

GPT-4.5 was the model that proved scaling GPT-4 on writing and emotional-intelligence benchmarks still produced meaningful gains, paving the way for GPT-5's unified "auto-switching" design. Even though GPT-4.5 was short-lived in ChatGPT (about 5 months as an active model), its contributions shaped the current lineup:

• Warmer conversational tone: GPT-4.5 introduced a more natural, less robotic voice that OpenAI carried into GPT-5.3's personality presets (Friendly, Professional, Candid, Quirky)
• Lower hallucination benchmark: GPT-4.5's 37% SimpleQA hallucination rate was a turning point; GPT-5.4 further reduced factual errors by 33% compared to GPT-5.2
• Research-preview model: GPT-4.5 was OpenAI's model for "knowledge tasks we can't solve with pure reasoning," and GPT-5.4 Pro now holds that role for high-stakes workflows
• Unified intelligence: GPT-4.5's awkward position alongside o-series reasoning models pushed OpenAI to unify everything under GPT-5's auto-switching router

Organize Your ChatGPT Conversations with ChatGPT Toolbox

ChatGPT Toolbox works with GPT-5.3 Instant, GPT-5.4 Thinking, GPT-5.4 Pro, and every other current ChatGPT model, adding folders, full-text search, bulk export, and a prompt library directly inside chatgpt.com. When you're running parallel conversations across Instant and Thinking modes, native ChatGPT gives you a single chronological sidebar with no categorization.

ChatGPT Toolbox's full-text search scans across all conversation names and message content, including conversations with any model. Free users get 2 folders, 2 pinned chats, 2 saved prompts, up to 5 search results per query, and the media gallery. Premium ($9.99/month or $99 lifetime) unlocks unlimited folders, unlimited full-text search, bulk export to TXT and JSON, prompt chaining, and cross-device sync.

For team deployments, the Enterprise plan ($12/seat/month or $10/seat/year, minimum 5 seats) adds admin dashboard, centralized billing, seat management, and team analytics. Also available for Google Gemini and Claude (from the makers of ChatGPT Toolbox).

Frequently Asked Questions

Is ChatGPT 4.5 still available in 2026?

No. ChatGPT 4.5 was removed from the OpenAI API on July 14, 2025 and from ChatGPT Plus, Team, and Enterprise on August 7, 2025 when GPT-5 launched. ChatGPT Pro users may find GPT-4.5 under the "Legacy Models" tab for a limited window, but it is not a practical option for new work. Current users should use GPT-5.3 Instant or GPT-5.4 Thinking instead.

What replaced ChatGPT 4.5?

GPT-5.3 Instant replaced GPT-4.5 for everyday writing and knowledge tasks - it's now the default model for all ChatGPT users, including free. GPT-5.4 Thinking replaced GPT-4.5 for research, long-form analysis, and tasks needing document understanding. GPT-5.4 Pro replaced GPT-4.5 for the hardest knowledge-work tasks and long-running workflows on Pro, Business, Enterprise, and Edu plans.

Was ChatGPT 4.5 the same as GPT-5?

No. GPT-4.5 (Orion) was OpenAI's last major GPT-4 series model and was positioned as a "research preview." GPT-5 launched on August 7, 2025 as a separate, more capable family that unified reasoning and general-purpose intelligence under a single auto-switching router. GPT-5 made GPT-4.5 redundant, which is why OpenAI retired it.

Can I still use GPT-4.5 via the OpenAI API?

No. GPT-4.5 was removed from the OpenAI API and Developer Playground on July 14, 2025. Developers with workflows built around GPT-4.5 should migrate to gpt-5.4 or gpt-5.3-instant, which handle GPT-4.5's use cases and outperform it on OpenAI's published benchmarks.

Why did OpenAI retire GPT-4.5 so quickly?

GPT-4.5 was only active for about 5 months as a mainstream ChatGPT option (February to August 2025). OpenAI retired it because GPT-5 surpassed it on writing quality, knowledge accuracy, and reasoning, while unifying everything under a single model picker. Maintaining GPT-4.5 as a separate option no longer provided value to users, and the API's premium pricing ($75/M input, $150/M output) made it economically unattractive for most developers.

What was the biggest difference between GPT-4.5 and GPT-4o?

GPT-4.5 had broader world knowledge, a lower hallucination rate (37% vs 62% on SimpleQA), and a more natural writing tone. GPT-4o was faster, multimodal (text, image, audio), and cheaper. GPT-4.5 was better for writing and research; GPT-4o was better for speed, voice, and visual tasks. Both are now retired - current users should use GPT-5.3 Instant instead.

Where can I read about the current ChatGPT models?

Read our complete ChatGPT models guide (2026) for a full breakdown of GPT-5.3 Instant, GPT-5.4 Thinking, GPT-5.4 Pro, GPT-5.2 Thinking (legacy), and specialized models. The OpenAI help center article on GPT-5.3 and GPT-5.4 is the authoritative source for model availability, context windows, and usage limits.

How do I organize ChatGPT conversations across different model modes?

Native ChatGPT doesn't group conversations by model or mode (Instant, Thinking, Pro). ChatGPT Toolbox adds folders, full-text search, and custom tags so you can keep GPT-5.3 Instant drafts separate from GPT-5.4 Thinking research threads. Premium ($9.99/month) unlocks unlimited folders and full-text search across all chats.

Bottom Line

ChatGPT 4.5 was a short-lived but influential research preview model from OpenAI that shaped the current GPT-5 line's writing tone, hallucination benchmarks, and unified-model design. As of April 2026, it is retired from the OpenAI API, ChatGPT Plus, Team, Business, and Free tiers. Only ChatGPT Pro users may find it via the Legacy Models tab, and only for a limited window. For any new work, use GPT-5.3 Instant (the new default for all users, free included) for writing and knowledge, GPT-5.4 Thinking for research and reasoning-heavy tasks, or GPT-5.4 Pro for the hardest long-running workflows. To keep conversations organized across models, install ChatGPT Toolbox (free, 20,000+ users, 4.5/5 rating).

Official OpenAI sources:

• GPT-5.3 and GPT-5.4 in ChatGPT (OpenAI Help Center)
• Retiring GPT-4o and older models (OpenAI, 2026)
• Introducing GPT-5.4 (OpenAI, March 2026)
• Introducing GPT-4.5 (OpenAI, February 2025)

Last updated: May 9, 2026