# What is a knowledge cutoff?

The date a model's training text ends. The model knows nothing after it unless it is given new text, for example by a web search.

- Page: https://tecyfy.com/concepts/knowledge-cutoff
- Subject: How AI works (https://tecyfy.com/concepts)
- Level: beginner
- Also called: Training Cutoff, Training Data Cutoff
- The lesson: How AI learns: a guessing game, not a database (https://tecyfy.com/learn/how-ai-works/how-ai-learns)
- Last checked: 2026-10-14

A model learns from a huge body of text collected up to a certain date. Then it is trained, tested and released. The date the text ends is the knowledge cutoff. Whatever happened after it, the model has never read about.

## The gap is longer than you think

The cutoff is always before the release, sometimes well before. 3 examples from models released in 2026:

| Model | Released | Knowledge cutoff |
|---|---|---|
| Claude Opus 5.5 | 22 September 2026 | June 2026 |
| GPT-6.1 Sol | 29 September 2026 | 30 April 2026 |
| Gemini 3.8 Flash | 2 September 2026 | March 2026 |

Google adds that in some subjects Gemini 3.8 Flash's knowledge ends in January 2025, over a year earlier than its headline cutoff. And a model stays in use for months after release, so by the time you ask, the gap can be a year or more.

## What it looks like

Ask about last month's news, this year's prices or the newest version of a tool, and a model without a search may answer with what was true at its cutoff, as if it were still true. It may not mention that it could be out of date. This is one road to a [Hallucination](https://tecyfy.com/concepts/hallucination).

Telling it the right answer helps for that conversation only. A model does not learn from your corrections. It answers from what is in front of it.

## How models answer past it

The usual fix is to give the model fresh text at the moment it answers. With web search, the service runs the searches and hands the results to the model, which can then answer beyond its cutoff and cite its sources. Pasting a document or connecting your files does the same for your own material. That is the idea behind [RAG](https://tecyfy.com/concepts/rag).

## What to do with this

For anything that changes, such as prices, versions, laws, people in a job or recent events, check that the answer used a search and cites a source with a date. If it did not, assume it may be out of date. You can find a model's cutoff on its maker's model page.

## Related

- [Hallucination](https://tecyfy.com/concepts/hallucination): When a model states something false as if it were true, such as a wrong date, a made-up quote or a source that does not exist.
- [RAG](https://tecyfy.com/concepts/rag): A way to answer from your own documents. A search finds the passages that match the question, and they are added to the prompt before the model writes.
- [Large Language Model](https://tecyfy.com/concepts/large-language-model): A model trained on a huge amount of text to predict the next token. Further training turns it into an assistant that follows instructions.

## Sources

- [Claude Opus 5.5](https://platform.claude.com/docs/en/models/opus-5-5/overview), Anthropic, Claude API documentation, 2026-10-14. Claude Opus 5.5 was released on 22 September 2026. Its training data cutoff and its reliable knowledge cutoff are both listed as June 2026.
- [GPT-6.1 Sol](https://developers.openai.com/api/docs/models/gpt-6.1-sol), OpenAI API documentation, 2026-10-14. GPT-6.1 Sol has a knowledge cutoff of 30 April 2026. OpenAI's API changelog dates its release to 29 September 2026.
- [Gemini 3.8 Flash model card](https://deepmind.google/models/model-cards/gemini-3-8-flash/), Google DeepMind, 2026-10-14. The knowledge cutoff of Gemini 3.8 Flash is March 2026, and in some domains its knowledge is limited to January 2025.
- [Gemini API release notes](https://ai.google.dev/gemini-api/docs/changelog), Google, 2026-10-14. Gemini 3.8 Flash became generally available on 2 September 2026.
- [Web search tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool), Anthropic, Claude API documentation, 2026-10-14. With web search, the service runs the searches and gives the results to the model, which answers beyond its knowledge cutoff and cites its sources.
- [Working memory](https://academy.claude.com/courses/ai-capabilities-and-limitations/working-memory), Anthropic Academy, 2026-10-14. The model does not learn from your corrections. It only responds to what is currently in its context.