94.0%

Support tickets, judged right.

Reflex is a custom decision model for support workflows. It returns typed answers and probabilities through one simple API.

Get an API key Read the API docs →

SemIf-144 accuracy: 78.5%
Queue routing (4-way): 95.3%
Priority scoring: 92.0%
Median, hosted: 317 ms

Reflex custom model benchmarks

Measured on queue routing, mood calls, priority scoring, SemIf-144, and Banking77-1200. Each result is one judgment per question.

Queue routing (4-way): 95.3%
Mood calls: 94.7%
Priority scoring: 92.0%
SemIf-144: 78.5%
Banking77-1200: 89.7%
Hosted median: 317 ms

Accuracy against Jev

Reflex Jev

OOD choice · 300 each

95.3%

89.0%

OOD yes/no · 300 each

94.7%

90.3%

OOD score · 300 each

92.0%

39.7%

SemIf-144 · 144 examples each

78.5%

97.2%

Banking77 · 1,200 examples each

89.7%

88.5%

Reflex and Jev were evaluated on the same stored examples for each comparison. Benchmarks use published evaluation sets and reserved test splits where available; results describe these tasks and do not guarantee performance on other data.

Two ways to watch it decide.

Reflex Pilot

A city car driven entirely by classifier calls — every steering choice streams from this API at ~300 ms. It drove a full 735 m route with zero contacts.

735 m · 0 contacts · J to engage

Play driving

2048

One board, one Reflex custom model call per move. It reached tile 64 in 74 moves with zero invalid answers. Play manually with arrows, or let it play.

74 moves · tile 64 · arrows to play

Play 2048

Tetris

One piece, one Reflex custom model call, one placement chosen from a shortlist. It cleared 5 lines across a 49-piece game in a real browser.

49 pieces · 5 lines · Space to drop

Play Tetris

Simulation, physics, game rules, and artwork © their authors; credits ship in each game and in the project’s attribution files.