Chat model

Qwen3.8-27B

stated confidence · reasoning off · 1 question per request

54.2% of the 2,198 questions right (52.1% to 56.3%, 95% interval). Back to the leaderboard.

How we called it

Kind: Chat model
Access: vLLM, self-hosted, BF16
Paid as: Our own hardware
Model ID: qwen38-27b-bf16
Variant: stated confidence
Parameters: 27.8B
Hardware: 1 × RTX PRO 6000 Blackwell (96 GB); BF16
Note: Open-Jev's base model
Confidence from: Written estimate
Run: 53, 4 October 2026

Settings

Reasoning: off
Request format: 1 question per request
Requests at once: 1
Temperature: not recorded
Harness seed: 0
Harness commit: not recorded
Prompt version: not recorded

Results

Questions	Accuracy	95% interval	Right
All questions	54.2%	52.1% to 56.3%	1,191 of 2,198
Reading	83.4%	80.3% to 86.0%	547 of 656
Grammar	41.8%	39.3% to 44.2%	644 of 1,542

Comparison group: BC. Compare with another setup.

Accuracy by topic

Topic	Part	Accuracy	Right
Word and phrase meanings	Reading	77.9%	102 of 131
Sentence meanings	Reading	86.4%	133 of 154
How a writer explains and develops ideas	Reading	74.3%	52 of 70
Topic and main point of a paragraph	Reading	92.9%	91 of 98
How a paragraph fits together	Reading	74.5%	79 of 106
Details that support a paragraph’s main point	Reading	92.8%	90 of 97
Word types, such as nouns and verbs	Grammar	39.3%	55 of 140
Groups of words built around a noun	Grammar	33.8%	22 of 65
Verbs and their forms	Grammar	45.1%	41 of 91
Verb forms used like nouns or modifiers	Grammar	38.9%	28 of 72
Word endings and what they do	Grammar	49.4%	39 of 79
How words are built	Grammar	27.5%	19 of 69
The roles of words and phrases in a sentence	Grammar	42.8%	62 of 145
How verbs show who acts and who receives the action	Grammar	45.9%	34 of 74
Types of sentences	Grammar	41.2%	35 of 85
Questions combining several grammar topics	Grammar	23.6%	52 of 220
Sound patterns and changes	Grammar	37.7%	29 of 77
Spelling	Grammar	48.1%	64 of 133
Punctuation	Grammar	50.4%	59 of 117
Problems in sentence wording	Grammar	60.0%	105 of 175

Confidence

Answers kept at 99% or more: 8, of which 1 wrong (share right: 52.9% to 97.8%, 95% interval); 2,190 skipped, the model less than 99% sure
Calibration gap: 40.8 points
AUROC, all questions: 0.51 0.51 to 0.52
AUROC, reading: 0.54 0.52 to 0.56
AUROC, grammar: 0.50 0.50 to 0.51
Brier score: 0.415
Accuracy of its surest answers: most confident fifth: 57.0% right; most confident half: 54.9% right; most confident four fifths: 54.4% right
Typical score of a right answer: 95.0%
Typical score of a wrong answer: 95.0%

Repeats

Run 5 times on all questions.

Answers changed in 5 runs: 44 of 2,198
Same wrong answer in all 5: 975 of 2,198
Wrong in at least one run: 1,019 of 2,198
Accuracy in each run: 54.2%, 54.3%, 54.3%, 54.3%, 54.3%

Latency

Latency, median request: 0.433 s (its repeated runs)
Latency, 90th percentile: 0.475 s
Requests at once: 1
Throughput: 147.0 questions per minute of request time
Request time: 15.0 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

$ per 1,000 questions: $0.23
Whole run: $0.52
Basis: Estimated at Alibaba's list price for the same model; our hardware cost is not measured.
Price: $0.50 input, $0.10 cached input, $3 output per 1M tokens (source, 2026-10-18)
Input tokens per question: 336
Output tokens per question: 22