Overview
GPT-5.5 Instant is OpenAI's fast, low-latency model that became ChatGPT's new default for all users on May 5, 2026, replacing GPT-5.3 Instant. It is the everyday workhorse of the GPT Instant line — the model most people actually talk to in ChatGPT — tuned for quick, conversational replies without the extended deliberation of OpenAI's separate thinking models. OpenAI summarized the release as 'smarter, clearer, and more personalized.'
Compared with GPT-5.3 Instant, the headline improvement is factuality: OpenAI reported that GPT-5.5 Instant produces 52.5% fewer hallucinated claims on high-stakes prompts covering medicine, law, and finance, while keeping the same instant response latency. It also posts large jumps on hard reasoning and multimodal evaluations — 81.2 on AIME 2025 (up from 65.4) and 76 on the MMMU-Pro multimodal reasoning benchmark (up from 69.2).
GPT-5.5 Instant accepts text and image input and returns text. It adds deeper personalization: with memory enabled it can refer back to your past conversations, uploaded files, and (when connected) Gmail to give more tailored answers, and a new 'memory sources' view shows which stored context shaped each response so you can edit or remove individual entries. For developers, OpenAI exposes the model through the API as the chat-latest alias rather than a pinned snapshot, so apps on chat-latest track the current default. As a member of the GPT-5.5 family it carries the same 1,050,000-token context window, up to 128,000 output tokens, and a December 1, 2025 knowledge cutoff.
| Released | 2026-05-05 |
|---|---|
| License | Proprietary |
| Weights | API only |
| Parameters | Undisclosed |
| Context | 1.05M |
| Max output | 128K |
| Architecture | Undisclosed (non-reasoning, low-latency) |
| Knowledge cutoff | 2025-12-01 |
| Modalities | Text, Vision |
| Status | Available |
Benchmarks
Benchmark scores OpenAI published for GPT-5.5 Instant against its predecessor GPT-5.3 Instant at launch (May 5, 2026).
| Benchmark | GPT-5.3 Instant | GPT-5.5 Instant |
|---|---|---|
| AIME 2025 | 65.4% | 81.2% |
| MMMU-Pro | 69.2% | 76% |
This model's scores
- AIME 2025 (math)81.2%
- MMMU-Pro (multimodal reasoning)76%
- Hallucination reduction vs GPT-5.3 Instant (high-stakes prompts)52.5%
- HealthBench (length-adjusted)51.4%
- HealthBench Hard (length-adjusted)22.9%
Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.
Pricing
| Input | $5.00 / 1M tokens |
|---|---|
| Cached input | $0.50 / 1M tokens |
| Output | $30.00 / 1M tokens |
GPT-5.5 family rate; Instant is served via the chat-latest alias with no separate Instant tier.
Strengths
- Much lower hallucination rate than GPT-5.3 Instant on high-stakes medical, legal, and financial prompts
- Instant, low-latency responses for everyday chat — no extended reasoning wait
- Strong multimodal reasoning over images and charts (MMMU-Pro 76)
- Built-in personalization from past conversations, files, and connected Gmail
- Very large 1,050,000-token context window for long documents and threads
- Free to use as the default ChatGPT model, including on the free tier
Best for
- Everyday ChatGPT conversation, drafting, and quick Q&A
- Higher-trust questions in medicine, law, and finance where factuality matters
- Image and chart understanding alongside text
- Personalized assistance grounded in your own history, files, and email
- Latency-sensitive chat apps that call the chat-latest API endpoint
How to access
| Provider | Model ID |
|---|---|
| OpenAI API ↗ | chat-latest |
GPT Instant — every version
The full lineage of the GPT Instant line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.
| Version | Released | Context | License |
|---|---|---|---|
| GPT-5.5 Instantcurrent | 2026-05-05 | — | Proprietary |
| GPT-5.3 Instant | 2026-03-03 | — | Proprietary |
| GPT-5.2 Instant | 2025-12-11 | — | Proprietary |
| GPT-5.1 Instant | 2025-11-12 | — | Proprietary |
FAQ
What is GPT-5.5 Instant and how is it different from GPT-5.5?
GPT-5.5 Instant is the fast, low-latency model in OpenAI's GPT Instant line that became ChatGPT's default for all users on May 5, 2026. It answers quickly without the extended deliberation of OpenAI's separate thinking models. It is part of the same GPT-5.5 family as the flagship GPT-5.5 (announced April 23, 2026) and shares its 1,050,000-token context window and December 1, 2025 knowledge cutoff, but it is tuned for speed and everyday conversation rather than long reasoning.
How much more accurate is GPT-5.5 Instant than GPT-5.3 Instant?
OpenAI reported that GPT-5.5 Instant produces 52.5% fewer hallucinated claims than GPT-5.3 Instant on high-stakes prompts covering medicine, law, and finance. It also improved on hard benchmarks, scoring 81.2 on AIME 2025 (up from 65.4) and 76 on MMMU-Pro multimodal reasoning (up from 69.2).
How do I use GPT-5.5 Instant through the API?
OpenAI serves GPT-5.5 Instant through the API as the chat-latest alias rather than a pinned snapshot, so requests to chat-latest track ChatGPT's current default model. As a GPT-5.5 family model it is priced at $5.00 per 1M input tokens and $30.00 per 1M output tokens, with cached input at $0.50 per 1M tokens; there is no separate Instant pricing tier.
Is GPT-5.5 Instant free to use?
Yes. GPT-5.5 Instant is ChatGPT's default model and is available to all users, including the free tier. Advanced personalization features that draw on past conversations, files, and connected Gmail rolled out first to Plus and Pro users on the web, with Free, Go, Business, and Enterprise tiers following over the weeks after launch.