Analytics

Public & anonymized. Same page for everyone, no account needed. Rows hold shape, never identity: no ip, no cookie id, no user id, no answer text. Cells under 7 observations stay hidden. How this data works · JSON

nu (pre-K through grade 12) · Last 28 days · computed 2026-10-19 07:59 UTC · refreshes every few minutes

4067 Model & service calls

5.2% Succeeded

0s Median latency

2.7 Calls per graded answer grading purposes only

742774 Characters sent

3294 Characters received

Every outbound call to a language model, the decision classifier, translation, speech, lesson generation or code execution. Purpose, host, latency & outcome only. No prompt or completion text.

By purpose

Purpose	Calls	Share	Succeeded	Median	95th	Chars in	Chars out
Text to speech	4036	99.2%	4.5%	0s	1s	599932	0
Classification (chat model)	11	0.3%	90.9%	1s	10s	66545	315
Feedback	10	0.2%	90.0%	1s	4s	44690	2054
Rubric scoring	6	0.1%	100.0%	1s	4s	31547	925
Code execution	4	0.1%	100.0%	2s	6s	60	0

By lesson

Which lessons drove model calls — chat, feedback and speech attributed to the lesson the learner was on.

Lesson	Subject	Calls	Share
Python: First Steps	computer_science	13	0.3%
ABCs: Letters and Sounds	language_arts	9	0.2%

By endpoint

Host	Calls	Share	Succeeded	Median
speech.ai.unturf.com	4036	99.2%	4.5%	0s
qwen.ai.unturf.com	22	0.5%	100.0%	1s
hermes.ai.unturf.com	5	0.1%	60.0%	1s
unsandbox	4	0.1%	100.0%	2s

By model

tts-1-f5 4036 (99.2%)

turboderp/Qwen3.8-27B-exl3 27 (0.7%)

python 4 (0.1%)

By brand

nu (pre-K through grade 12) 4067 (100.0%)

Calls over time

Failed calls over time

Calls over time by purpose

Calls: classify

Calls: feedback

Calls: rubric

Calls: tts

Failure reasons

HTTPError 3839

HTTP 500 11

ReadTimeout 5

unavailable 1

HTTP 429 1