# Jev set the standard.

Jev is the model behind ThinkThen. TypeSafe makes it. It reads a text and a question and returns a probability for every answer. It writes no text, and you train nothing.

Roger Bannister's mile set a standard for runners. Jev answers in about 60 milliseconds. That is the median time at Jev's server, measured on 2026-09-26. On Beatles Bench, a thousand answers cost 0.016 dollars.

The numbers come from [the bench's record for this slide](https://github.com/botassembly/beatles-bench/tree/main/examples/jev).

## Run it

*At a bar of 0.8, Jev picks Ringo.*

```
is_a_song="The text is the title of a song by the Beatles."
question="$is_a_song Who sings the lead vocal on it?"
singers=(
  John
  Paul
  George
  Ringo
  "John and Paul duet"
)

printf '%s' "Octopus's Garden" |
thinkthen choose "$question" "${singers[@]}" \
  --threshold 0.8 \
  --replay recording
```

*Output*

```
"Ringo"
```

*exit 0: an option came back*

## Change the bar

*A bar of 0.9 asks for more. The answer is not sure.*

```
is_a_song="The text is the title of a song by the Beatles."
question="$is_a_song Who sings the lead vocal on it?"
singers=(
  John
  Paul
  George
  Ringo
  "John and Paul duet"
)

printf '%s' "Octopus's Garden" |
thinkthen choose "$question" "${singers[@]}" \
  --threshold 0.9 \
  --replay recording
```

*Output*

```
null
```

*exit 3: not sure*

## The lesson

Ringo is right. Jev's probability for Ringo clears 0.8 and falls short of 0.9. A bar of 0.9 asks for more than Jev gives.

Jev answers with a probability, and your bar decides what counts.

On GitHub: [github.com/botassembly/beatles-bench/tree/main/examples/choose](https://github.com/botassembly/beatles-bench/tree/main/examples/choose)