How is this different from a plagiarism or text-compare tool?

Plagiarism and diff tools compare word overlap - the literal characters that match. This calculator compares meaning using AI embeddings, so paraphrased text with no shared words can still score as highly similar.

How is the similarity score calculated?

Each text is converted into an embedding vector with OpenAI's text-embedding-3-small model. The cosine similarity between the two vectors is then rescaled to an intuitive 0-100% score.

Is the semantic similarity calculator free?

Yes. You can compare two texts and get a semantic similarity score for free, with no signup required. Creating an account is optional and only adds saved comparison history.

What happens to the text I enter?

It is sent to OpenAI for the embeddings and kept on this site, with the score, so we can see how the tool is used and improve it. Please do not enter personal or confidential information.

Jev is TypeSafe AI's System One decision model. Instead of generating prose it takes a state (text or JSON) and typed questions - a Noul (a yes/no question, answered as the probability of yes), a Choice between options, or a Score on a scale - and returns calibrated probabilities for each answer.

Yes while the sponsored pool lasts. Each run costs a fraction of a cent in input tokens (USD 0.042 per million) and is paid from a USD 10 pool. When it is spent, Try Jev pauses.

What happens to my text in Try Jev?

The state and question are forwarded to api.typesafe.ai for the answer and kept on this site, with the answer, so we can see how Jev is used and improve the examples. Please do not enter personal or confidential information.

Try Jev

Ask TypeSafe's Jev a question about any text or JSON. It answers with probabilities, not prose.

Example

State

What is Jev?

Jev is TypeSafe AI's System One model: a decision model rather than a chat model. You hand it a state (free text or a JSON object) and one or more typed questions, and it returns a probability for each answer instead of a paragraph. That makes the output something code can branch on directly.

Three question types

noul is a yes/no question, or a statement to judge, and comes back as one number: the probability that the answer is yes, 0 to 1. That number is the answer and the certainty in one, so there is no separate confidence; near 0.5 means the model gives yes and no similar odds, and the threshold that turns it into a decision lives in your code. It is not a measure of degree: for that, ask a score. choice picks one of up to 255 named options and returns the full distribution plus a confidence. score rates the state on a scale of 2 to 10 described levels and returns the expected level with the distribution behind it. Every question can carry short criteria that pin down what the answers mean.

Frequently asked questions

Is trying Jev free?: Yes while the sponsored pool lasts. Jev charges for input tokens only, USD 0.042 per million, so a typical run here costs well under a hundredth of a cent and comes out of a USD 10 pool. When the pool is spent, Try Jev pauses.
What happens to my text?: The state and question are forwarded to api.typesafe.ai for the answer and kept on this site, with the answer, so we can see how Jev is used and improve the examples. Please do not enter personal or confidential information. Identical runs are answered from a short-lived in-memory cache so repeat clicks cost nothing.
How do I use the API myself?: Run any example and open "The request we sent": the JSON and a ready-to-paste curl command are exactly what this page sent. Add a key from console.typesafe.ai and you have a working integration.

Semantic Similarity Calculator

Paste two texts and get a 0 to 100% score for how close they are in meaning, not in wording.

Text A

Text B

What is semantic similarity?

Semantic similarity measures how close two pieces of text are in meaning, rather than how many words they have in common. Two sentences can share almost no words and still mean the same thing - and this calculator is built to catch exactly that.

Meaning, not word overlap

Plagiarism checkers and text-compare (diff) tools look at literal word overlap. This semantic similarity calculator compares meaning using AI embeddings, so a paraphrase with completely different wording can still score as highly similar. That makes it useful for comparing paraphrases, deduplicating content, matching questions to answers, evaluating model outputs, and checking whether two passages say the same thing.

How the similarity score is calculated

Each text is converted into an embedding vector using OpenAI's text-embedding-3-small model. We then take the cosine similarity between the two vectors and rescale it to an intuitive 0-100% score - higher means closer in meaning.

Frequently asked questions

Is the semantic similarity calculator free?: Yes. You can compare two texts and get a similarity score for free, with no signup required. An account is optional and only adds saved comparison history.
How is this different from a plagiarism or text-compare tool?: Those tools measure word overlap. This one measures meaning, so paraphrased text with no shared words can still score as highly similar.
What is cosine similarity?: Cosine similarity is the cosine of the angle between two vectors. For text embeddings it captures how aligned two meanings are, independent of length.
What happens to the text I enter?: It is sent to OpenAI for the embeddings and kept on this site, with the score, so we can see how the tool is used and improve it. Please do not enter personal or confidential information.

Welcome back

Try Jev

Three question types

Frequently asked questions

Semantic Similarity Calculator

Meaning, not word overlap

How the similarity score is calculated

Frequently asked questions