Decision model
Perplexity Decider Flash v1.2
local · 1 question per request
47.8% of the 2,198 questions right (45.7% to 49.9%, 95% interval). Back to the leaderboard.
How we called it
- Kind
- Decision model
- Access
- Perplexity Decider server, self-hosted
- Paid as
- Our own hardware
- Model ID
pplx-decider-flash-v1.2- Variant
- local
- Parameters
- 4.5B
- Hardware
- 1 × DGX Spark, GB10 (128 GB unified memory); BF16
- Note
- Perplexity's released Flash checkpoint, based on Qwen3.5-4B, run on our hardware. An API run is pending; this is an additional local setup.
- Confidence from
- Option score
- Run
- 223, 17 October 2026
Settings
- Reasoning
- off
- Request format
- 1 question per request
- Requests at once
- 1
- Temperature
- not recorded
- Harness seed
- 0
- Harness commit
f9f71e0fa923- Prompt version
254597b8be16
Results
| Questions | Accuracy | 95% interval | Right |
|---|---|---|---|
| All questions | 47.8% | 45.7% to 49.9% | 1,050 of 2,198 |
| Reading | 72.7% | 69.2% to 76.0% | 477 of 656 |
| Grammar | 37.2% | 34.8% to 39.6% | 573 of 1,542 |
Comparison group: DE. Compare with another setup.
Accuracy by topic
Confidence
- Answers kept at 99% or more
- 0, of which 0 wrong; 2,198 skipped, the model less than 99% sure
- Calibration gap
- 8.3 points
- AUROC, all questions
- 0.76 0.74 to 0.78
- AUROC, reading
- 0.84 0.81 to 0.87
- AUROC, grammar
- 0.66 0.63 to 0.69
- Brier score
- 0.203
- Accuracy of its surest answers
- most confident fifth: 88.2% right; most confident half: 65.4% right; most confident four fifths: 53.7% right
- Typical score of a right answer
- 68.9%
- Typical score of a wrong answer
- 41.5%
Repeats
Run 5 times on all questions.
- Answers changed in 5 runs
- 0 of 2,198
- Same wrong answer in all 5
- 1,148 of 2,198
- Wrong in at least one run
- 1,148 of 2,198
- Accuracy in each run
- 47.8%, 47.8%, 47.8%, 47.8%, 47.8%
Latency
- Latency, median request
- 0.132 s (this run)
- Latency, 90th percentile
- 0.187 s
- Requests at once
- 1
- Throughput
- 419.9 questions per minute of request time
- Request time
- 5.2 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
- $0.004
- Whole run
- $0.009
- Basis
- Estimated at Perplexity's list price for the same model; our hardware cost is not measured.
- Price
- $0.01 input, output free per 1M tokens, plus 3 billed input tokens per request (source, 2026-10-18)
- Input tokens per question
- 385
- 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 | 7.3% | −0.09 points | 7%§ |
| GPT-6 Astrahigh reasoning · 16 questions per request | 7.4% | −0.10 points | 7%§ |
| GPT-6.1 Solhigh reasoning · 16 questions per request | 7.4% | −0.10 points | 7%§ |
| GPT-6 Astralow reasoning · 16 questions per request | 7.8% | −0.14 points | 8%§ |
| GPT-6 Astralow reasoning · 1 question per request | 7.3% | −0.09 points | 7%§ |
| Gemini 3.8 Flashlow reasoning · 1 question per request | 7.3% | −0.09 points | 7%§ |
| GPT-6.1 Sollow reasoning · 1 question per request | 7.4% | −0.10 points | 7%§ |
| GPT-6.1 Sollow reasoning · 16 questions per request | 7.8% | −0.14 points | 7%§ |
| Claude Opus 5.5high reasoning · 16 questions per request | 7.7% | −0.13 points | 8%§ |
| GPT-6.1 Terrahigh reasoning · 16 questions per request | 7.4% | −0.10 points | 7%§ |
| Claude Opus 5.5low reasoning · 1 question per request | 7.8% | −0.14 points | 8%§ |
| Gemini 3.8 Flashlow reasoning · 16 questions per request | 7.7% | −0.13 points | 6%§ |
| GPT-6.1 Terralow reasoning · 16 questions per request | 7.4% | −0.10 points | 6%§ |
| Claude Opus 5.5low reasoning · 16 questions per request | 7.8% | −0.14 points | 8%§ |
| Claude Sonnet 5.5high reasoning · 1 question per request | 8.6% | −0.15 points | 9%§ |
| Claude Sonnet 5.5high reasoning · 16 questions per request | 11.7% | −0.14 points | 12%§ |
§ Counts the first model at its maker’s list price for the same model; our hardware cost is not measured.