Decision model

KnowLine-12B-Gen1

1 question per request

Accuracy, all 2,198 questions
62.5%60.4% to 64.5%, 95% interval; 1,373 right
Answers kept at 99% confidence
No answers at this thresholdnone of the 2,198 questions at 99% or more
Cost, $ per 1,000 questionsper 1M questions
$0.052$52.28Basis and price under Cost
Latency
0.071 smedian request, 1 question per request

Back to the leaderboard.

How we called it

Kind
Decision model
Access
KnowLine server (knowline_server.py over SGLang, PelaAI's release code), self-hosted
Paid as
Our own hardware
Model ID
PelaAI/KnowLine-12B-Gen1
Parameters
12B dense, as the maker states it (counted from the released files: 11.96B).the whole checkpoint, its vision encoder included (text only here); the card says 12B parameters; source: counted from the checkpoint's safetensors, PelaAI/KnowLine-12B-Gen1@f26fbc2, huggingface.co/PelaAI/KnowLine-12B-Gen1; read 15 November 2026
Hardware
1 × RTX PRO 6000 Blackwell (96 GB); BF16
Note
PelaAI's released checkpoint (13 November 2026), a fine-tune of Gemma-4-12B-it, run on our hardware through its card's serve_knowline.sh (SGLang 0.5.23 in the lmsysorg/sglang image; the card's launcher names 0.5.21, then knowline_server.py), text only.
Confidence from
Option score
Run
1309, 16 November 2026

Settings

Reasoning
off
Request format
1 question per request
Requests at once
1
Temperature
not recorded
Run record: harness seed, harness commit and prompt version
Harness seed
0
Harness commit
88e48d7a1055
Prompt version
254597b8be16

Results

QuestionsAccuracy95% intervalRight
All questions62.5%60.4% to 64.5%1,373 of 2,198
Reading86.0%83.1% to 88.4%564 of 656
Grammar52.5%50.0% to 54.9%809 of 1,542

Comparison group: KLMNOPQ. Compare with another setup.

Accuracy by topic
TopicPartAccuracyRight
Word and phrase meaningsReading81.7%107 of 131
Sentence meaningsReading94.2%145 of 154
How a writer explains and develops ideasReading80.0%56 of 70
Topic and main point of a paragraphReading93.9%92 of 98
How a paragraph fits togetherReading72.6%77 of 106
Details that support a paragraph’s main pointReading89.7%87 of 97
Word types, such as nouns and verbsGrammar53.6%75 of 140
Groups of words built around a nounGrammar52.3%34 of 65
Verbs and their formsGrammar71.4%65 of 91
Verb forms used like nouns or modifiersGrammar65.3%47 of 72
Word endings and what they doGrammar54.4%43 of 79
How words are builtGrammar36.2%25 of 69
The roles of words and phrases in a sentenceGrammar49.7%72 of 145
How verbs show who acts and who receives the actionGrammar60.8%45 of 74
Types of sentencesGrammar50.6%43 of 85
Questions combining several grammar topicsGrammar40.5%89 of 220
Sound patterns and changesGrammar39.0%30 of 77
SpellingGrammar54.1%72 of 133
PunctuationGrammar49.6%58 of 117
Problems in sentence wordingGrammar63.4%111 of 175

English

The Text tab's 547 kept reading questions translated into English, asked with this setup's Text tab settings. Every setup: English reading; against Turkish: English and Turkish.

Right answers
474 of 547, first English run (1313)
Accuracy
86.7%
Non-hallucination with “I don't know”
1.3%: 1 “I don't know” of 77 not right (run 1316)
Answered with “I don't know”
99.8%: 546 of 547
Error rate with “I don't know”
13.9%: 76 of 546 answers
Cost, $ per 1,000 questionsper 1M questions, first English run
0.05756.86from GPU time
Latency, median request, first English run
0.078 s; 1 request at once
English run
1313
Machine, English runs
RTX PRO 6000
Date
16 November 2026
Request format
1 question per request
Reasoning
off
Prompt version
47f5776e71db
Translations, SHA-256
bf0d47cd385a93050c65906977e199ef7e45d5848dbbdef2bc8d124f4d5abf68

Confidence

Answers kept at 99% or more
No answers at this threshold; 2,198 skipped, the model less than 99% sure
Calibration gap
6.2 points
AUROC, all questions
0.81 0.79 to 0.83
AUROC, reading
0.91 0.88 to 0.93
AUROC, grammar
0.73 0.71 to 0.76
Brier score
0.174
Accuracy of its surest answers
most confident fifth: 98.2% right; most confident half: 86.0% right; most confident four fifths: 70.0% right
Typical score of a right answer
67.6%
Typical score of a wrong answer
38.8%

Repeats

We did not repeat this setup.

Latency

Latency, median request
0.071 s (this run)
Latency, 90th percentile
0.097 s
Requests at once
1
Throughput
793.8 questions per minute of request time
Request time
2.8 minutes, over the requests whose answers were kept

Latency: the time from sending one request to receiving its answer, measured from our computer, network and any retries within the request included. Request time: the time at least one request was in flight, over the requests whose answers were kept. It is not the run's elapsed time: it leaves out requests that failed and any pause before a run was resumed.

Cost

Cost per 1,000 questionsper 1M questions
$0.052$52.28
Whole run
$0.11
Basis
Estimated from GPU time: the run's request time x the GPUs its server held x Runpod's on-demand price for an RTX PRO 6000, $2.49 per GPU-hour (read 5 November 2026). A cloud-equivalent estimate, not measured spending.
Price
$2.49 per GPU-hour, 1 GPU, for the run's request time (source, 2026-11-05)
Input tokens per question
737
Output tokens per question
1