Chat model

GPT-6 Astra

low reasoning · 1 question per request

97.4% of the 2,198 questions right (96.7% to 98.0%, 95% interval). Back to the leaderboard.

We also ran this setup at 16 questions per request: GPT-6 Astra · low reasoning · 16 questions per request.

Which decision model can go first in front of it? See the Routing tab.

How we called it

Kind: Chat model
Access: Codex CLI, ChatGPT subscription
Paid as: Subscription
Model ID: gpt-6-astra
Confidence from: Written estimate
Run: 191, 13 October 2026

Settings

Reasoning: low
Request format: 1 question per request
Requests at once: 1
Temperature: not recorded
Harness seed: 0
Harness commit: 54043fd92cd5
Prompt version: 254597b8be16

Reasoning effort levels (OpenAI's model page, read 28 October 2026): low, medium, high, xhigh, max. Top level: max. This run: low.

Results

Questions	Accuracy	95% interval	Right
All questions	97.4%	96.7% to 98.0%	2,141 of 2,198
Reading	97.3%	95.7% to 98.3%	638 of 656
Grammar	97.5%	96.6% to 98.1%	1,503 of 1,542

Comparison group: abcd. Compare with another setup.

Accuracy by topic

Topic	Part	Accuracy	Right
Word and phrase meanings	Reading	97.7%	128 of 131
Sentence meanings	Reading	98.7%	152 of 154
How a writer explains and develops ideas	Reading	94.3%	66 of 70
Topic and main point of a paragraph	Reading	98.0%	96 of 98
How a paragraph fits together	Reading	95.3%	101 of 106
Details that support a paragraph’s main point	Reading	97.9%	95 of 97
Word types, such as nouns and verbs	Grammar	97.9%	137 of 140
Groups of words built around a noun	Grammar	96.9%	63 of 65
Verbs and their forms	Grammar	100.0%	91 of 91
Verb forms used like nouns or modifiers	Grammar	97.2%	70 of 72
Word endings and what they do	Grammar	98.7%	78 of 79
How words are built	Grammar	95.7%	66 of 69
The roles of words and phrases in a sentence	Grammar	97.2%	141 of 145
How verbs show who acts and who receives the action	Grammar	98.6%	73 of 74
Types of sentences	Grammar	98.8%	84 of 85
Questions combining several grammar topics	Grammar	95.9%	211 of 220
Sound patterns and changes	Grammar	97.4%	75 of 77
Spelling	Grammar	100.0%	133 of 133
Punctuation	Grammar	93.2%	109 of 117
Problems in sentence wording	Grammar	98.3%	172 of 175

Confidence

Answers kept at 99% or more: 1,839, of which 15 wrong (share right: 98.7% to 99.5%, 95% interval); 359 skipped, the model less than 99% sure
Calibration gap: 1.3 points
AUROC, all questions: 0.86 0.81 to 0.91
AUROC, reading: 0.93 0.87 to 0.97
AUROC, grammar: 0.82 0.76 to 0.88
Brier score: 0.022
Accuracy of its surest answers: most confident fifth: 100.0% right; most confident half: 99.5% right; most confident four fifths: 99.2% right
Typical score of a right answer: 99.0%
Typical score of a wrong answer: 97.0%

Repeats

We did not repeat this setup.

Latency

Latency, median request: 6.41 s (this run)
Latency, 90th percentile: 9.22 s
Requests at once: 1
Throughput: 8.6 questions per minute of request time
Request time: 257.0 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: $26.74
Whole run: $58.77
Basis: Subscription run priced as if we had used the API
Price: $10 input, $1 cached input, $12.50 cache writes, $50 output per 1M tokens (source, 2026-10-08)
Input tokens per question: 4,641
Output tokens per question: 59