Chat model

Claude Sonnet 5.5

high reasoning · 16 questions per request

85.0% of the 2,198 questions right (83.5% to 86.5%, 95% interval). Back to the leaderboard.

We also ran this setup at 1 question per request: Claude Sonnet 5.5 · high reasoning · 1 question per request.

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

How we called it

Kind: Chat model
Access: Claude Code CLI, Claude subscription
Paid as: Subscription
Model ID: claude-sonnet-5-5
Confidence from: Written estimate
Run: 55, 4 October 2026

Settings

Reasoning: high
Request format: 16 questions per request
Requests at once: 1
Temperature: not recorded
Harness seed: 0
Harness commit: not recorded
Prompt version: not recorded

Effort levels (Anthropic's effort documentation, read 28 October 2026): low, medium, high (default), xhigh, max. Top level: max. This run: high.

Results

Questions	Accuracy	95% interval	Right
All questions	85.0%	83.5% to 86.5%	1,869 of 2,198
Reading	85.2%	82.3% to 87.7%	559 of 656
Grammar	85.0%	83.1% to 86.7%	1,310 of 1,542

Comparison group: l. Compare with another setup.

Accuracy by topic

Topic	Part	Accuracy	Right
Word and phrase meanings	Reading	80.2%	105 of 131
Sentence meanings	Reading	89.6%	138 of 154
How a writer explains and develops ideas	Reading	68.6%	48 of 70
Topic and main point of a paragraph	Reading	92.9%	91 of 98
How a paragraph fits together	Reading	82.1%	87 of 106
Details that support a paragraph’s main point	Reading	92.8%	90 of 97
Word types, such as nouns and verbs	Grammar	93.6%	131 of 140
Groups of words built around a noun	Grammar	90.8%	59 of 65
Verbs and their forms	Grammar	86.8%	79 of 91
Verb forms used like nouns or modifiers	Grammar	94.4%	68 of 72
Word endings and what they do	Grammar	94.9%	75 of 79
How words are built	Grammar	84.1%	58 of 69
The roles of words and phrases in a sentence	Grammar	89.7%	130 of 145
How verbs show who acts and who receives the action	Grammar	91.9%	68 of 74
Types of sentences	Grammar	90.6%	77 of 85
Questions combining several grammar topics	Grammar	77.3%	170 of 220
Sound patterns and changes	Grammar	79.2%	61 of 77
Spelling	Grammar	75.9%	101 of 133
Punctuation	Grammar	80.3%	94 of 117
Problems in sentence wording	Grammar	79.4%	139 of 175

Confidence

Answers kept at 99% or more: 0, of which 0 wrong; 2,198 skipped, the model less than 99% sure
Calibration gap: 16.0 points
AUROC, all questions: 0.78 0.75 to 0.81
AUROC, reading: 0.81 0.76 to 0.86
AUROC, grammar: 0.77 0.74 to 0.80
Brier score: 0.135
Accuracy of its surest answers: most confident fifth: 96.0% right; most confident half: 94.8% right; most confident four fifths: 91.5% right
Typical score of a right answer: 80.0%
Typical score of a wrong answer: 50.0%

Repeats

Run 5 times on a sample of 200 questions.

Answers changed in 5 runs: 31 of 200
Same wrong answer in all 5: 9 of 200
Wrong in at least one run: 40 of 200
Accuracy in each run: 89.5%, 91.0%, 89.5%, 87.0%, 87.0%

Latency

Latency, median request: 20.5 s, measured on a sample (200 questions) over 5 runs
Latency, 90th percentile: 37.0 s
Requests at once: 1
Throughput: 37.8 questions per minute of request time
Request time: 58.5 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: $3.65
Whole run: $8.08
Basis: Subscription run priced as if we had used the API
Price: $2 input, $0.10 cached input, $2.50 5-minute cache writes, $4 1-hour cache writes, $10 output per 1M tokens (source, 2026-10-08)
Input tokens per question: 550
Output tokens per question: 199