Decision model
Liquid d1-ultra
1 question per request
77.1% of the 2,198 questions right (75.3% to 78.8%, 95% interval). Back to the leaderboard.
How we called it
- Kind
- Decision model
- Access
- Liquid AI decisions API
- Paid as
- Provider service
- Model ID
d1-ultra- Confidence from
- Option score
- Run
- 199, 16 October 2026
Settings
- Reasoning
- off
- Request format
- 1 question per request
- Requests at once
- 1
- Temperature
- not recorded
- Harness seed
- 0
- Harness commit
997f1d0fb734- Prompt version
254597b8be16
Results
| Questions | Accuracy | 95% interval | Right |
|---|---|---|---|
| All questions | 77.1% | 75.3% to 78.8% | 1,694 of 2,198 |
| Reading | 92.1% | 89.8% to 93.9% | 604 of 656 |
| Grammar | 70.7% | 68.4% to 72.9% | 1,090 of 1,542 |
Comparison group: mn. Compare with another setup.
Accuracy by topic
Page images (preliminary)
Preliminary: these numbers are not final and may change. This setup is condition A of the Page images tab, which asks the 763 questions with underlined or numbered words as text (A), as a page image (B), as both (C) and through OCR models (D); the share answered right, with its 95% interval.
| Model | A · text onlyaccuracy, higher is better | B · image onlyaccuracy, higher is better | C · text plus imageaccuracy, higher is better | Best OCR routeaccuracy, higher is better |
|---|---|---|---|---|
| Liquid d1-ultra1 question per request | 73.5%70.3% to 76.5% | 50.1%46.5% to 53.6% | 72.1%68.8% to 75.2% | 67.1%via Surya OCR 263.7% to 70.3% |
Confidence
- Answers kept at 99% or more
- 90, of which 0 wrong (share right: 95.9% to 100.0%, 95% interval); 2,108 skipped, the model less than 99% sure
- Calibration gap
- 3.6 points
- AUROC, all questions
- 0.85 0.83 to 0.87
- AUROC, reading
- 0.91 0.88 to 0.93
- AUROC, grammar
- 0.81 0.79 to 0.83
- Brier score
- 0.129
- Accuracy of its surest answers
- most confident fifth: 100.0% right; most confident half: 95.7% right; most confident four fifths: 85.9% right
- Typical score of a right answer
- 88.4%
- Typical score of a wrong answer
- 47.5%
Repeats
Run 5 times on all questions.
- Answers changed in 5 runs
- 375 of 2,198
- Same wrong answer in all 5
- 294 of 2,198
- Wrong in at least one run
- 669 of 2,198
- Accuracy in each run
- 77.1%, 76.7%, 75.9%, 76.7%, 76.0%
Latency
- Latency, median request
- 0.223 s (this run)
- Latency, 90th percentile
- 0.351 s
- Requests at once
- 1
- Throughput
- 242.0 questions per minute of request time
- Request time
- 9.1 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
- $0.014
- Whole run
- $0.032
- Basis
- Estimated from published prices and recorded token usage
- Price
- $0.042 input, output free per 1M tokens (source, 2026-10-16)
- Input tokens per question
- 341
- Output tokens per question
- 0
Going first in front of an LLM
Replay of recorded answers, cutoff chosen on half the questions, averaged over the splits. Savings are estimates from API prices; the Method page explains how.
| LLM behind it | Share taken | Accuracy change | Saving |
|---|---|---|---|
| Gemini 3.8 Flashhigh reasoning · 16 questions per request | 34.0% | −0.09 points | 33% |
| GPT-6 Astrahigh reasoning · 16 questions per request | 34.2% | −0.06 points | 34% |
| GPT-6.1 Solhigh reasoning · 16 questions per request | 34.2% | −0.07 points | 33% |
| GPT-6 Astralow reasoning · 16 questions per request | 34.2% | −0.06 points | 34% |
| GPT-6 Astralow reasoning · 1 question per request | 33.4% | −0.07 points | 33% |
| Gemini 3.8 Flashlow reasoning · 1 question per request | 34.3% | −0.12 points | 33% |
| GPT-6.1 Sollow reasoning · 1 question per request | 35.3% | −0.08 points | 35% |
| GPT-6.1 Sollow reasoning · 16 questions per request | 34.3% | −0.07 points | 33% |
| Claude Opus 5.5high reasoning · 16 questions per request | 36.8% | −0.06 points | 37% |
| GPT-6.1 Terrahigh reasoning · 16 questions per request | 34.7% | −0.06 points | 32% |
| Claude Opus 5.5low reasoning · 1 question per request | 37.8% | −0.10 points | 38% |
| Gemini 3.8 Flashlow reasoning · 16 questions per request | 33.1% | −0.11 points | 28% |
| GPT-6.1 Terralow reasoning · 16 questions per request | 36.2% | −0.08 points | 33% |
| Claude Opus 5.5low reasoning · 16 questions per request | 45.1% | −0.17 points | 45% |
| Claude Sonnet 5.5high reasoning · 1 question per request | 46.3% | −0.32 points | 46% |
| Claude Sonnet 5.5high reasoning · 16 questions per request | 75.4% | −0.07 points | 75% |