Decision model
d1.2
1 question per request
- Accuracy, all 2,198 questions
- 61.7%59.7% to 63.7%, 95% interval; 1,357 right
- Answers kept at 99% confidence
- 33 of 2,1980 wrong; share right 89.6% to 100.0%, 95% interval
- Cost, $ per 1,000 questionsper 1M questions
- $0.007$6.83Basis and price under Cost
- Latency
- 0.213 smedian request, 1 question per request
How we called it
- Kind
- Decision model
- Access
- Liquid AI decisions API
- Paid as
- Provider service
- Model ID
d1.2- Parameters
- not disclosed by makerpages checked: www.liquid.ai/blog/d1-2; read 16 November 2026
- Confidence from
- Option score
- Run
- 1346, 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
0f7123a34772- Prompt version
254597b8be16
Results
| Questions | Accuracy | 95% interval | Right |
|---|---|---|---|
| All questions | 61.7% | 59.7% to 63.7% | 1,357 of 2,198 |
| Reading | 86.0% | 83.1% to 88.4% | 564 of 656 |
| Grammar | 51.4% | 48.9% to 53.9% | 793 of 1,542 |
Comparison group: LMNOPQR. Compare with another setup.
Accuracy by topic
Vision (preliminary)
Preliminary: these numbers are not final and may change. This setup is condition A of the Vision 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, its 95% interval in the number's tooltip.
| 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 |
|---|---|---|---|---|
| d1.2 Closed · Efficient | 58.1% | 46.9% | 59.1% | 51.9%via Surya OCR 2 |
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
- 487 of 547, first English run (1357)
- Accuracy
- 89.0%
- Non-hallucination with “I don't know”
- 61.5%: 59 “I don't know” of 96 not right (run 1361)
- Answered with “I don't know”
- 89.2%: 488 of 547
- Error rate with “I don't know”
- 7.6%: 37 of 488 answers
- Cost, $ per 1,000 questionsper 1M questions, first English run
- 0.0076.66at list prices
- Latency, median request, first English run
- 0.219 s; 1 request at once
- English run
- 1357
- Machine, English runs
- hosted API
- 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
- 33, of which 0 wrong (share right: 89.6% to 100.0%, 95% interval); 2,165 skipped, the model less than 99% sure
- Calibration gap
- 3.3 points
- AUROC, all questions
- 0.81 0.79 to 0.82
- AUROC, reading
- 0.88 0.85 to 0.91
- AUROC, grammar
- 0.73 0.71 to 0.76
- Brier score
- 0.172
- Accuracy of its surest answers
- most confident fifth: 98.4% right; most confident half: 84.8% right; most confident four fifths: 68.7% right
- Typical score of a right answer
- 73.9%
- Typical score of a wrong answer
- 39.3%
Repeats
Run 5 times on all questions.
- Answers changed in 5 runs
- 706 of 2,198
- Same wrong answer in all 5
- 385 of 2,198
- Wrong in at least one run
- 1,091 of 2,198
- Accuracy in each run
- 61.7%, 60.9%, 61.7%, 61.3%, 60.6%
Latency
- Latency, median request
- 0.213 s (this run)
- Latency, 90th percentile
- 0.320 s
- Requests at once
- 1
- Throughput
- 242.5 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
- Cost per 1,000 questionsper 1M questions
- $0.007$6.83
- Whole run
- $0.015
- Basis
- Estimated from published prices and recorded token usage
- Price
- $0.02 input, output free per 1M tokens (source, 2026-11-16)
- 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 Flash high · chat | 12.9% | −0.10 points | 13% |
| GPT-6 Astra high · chat | 13.6% | −0.13 points | 14% |
| GPT-6.1 Sol high · chat | 13.7% | −0.13 points | 13% |
| GPT-6 Astra low · chat | 13.6% | −0.13 points | 13% |
| GPT-6 Astra low · chat · 1 per request | 12.2% | −0.10 points | 12% |
| Gemini 3.8 Flash low · chat · 1 per request | 15.4% | −0.17 points | 15% |
| GPT-6.1 Sol low · chat · 1 per request | 16.9% | −0.18 points | 17% |
| GPT-6.1 Sol low · chat | 13.7% | −0.13 points | 13% |
| Claude Opus 5.5 high · chat | 16.2% | −0.11 points | 16% |
| GPT-6.1 Terra high · chat | 15.3% | −0.16 points | 14% |
| Claude Opus 5.5 low · chat · 1 per request | 16.4% | −0.14 points | 16% |
| Gemini 3.8 Flash low · chat | 16.9% | −0.17 points | 14% |
| GPT-6.1 Terra low · chat | 15.1% | −0.15 points | 13% |
| Claude Opus 5.5 low · chat | 16.9% | −0.18 points | 17% |
| Claude Sonnet 5.5 high · chat · 1 per request | 20.7% | −0.25 points | 21% |
| Claude Sonnet 5.5 high · chat | 34.0% | −0.23 points | 34% |