Decision model
GLiDE 0.4B
1 question per request
- Accuracy, all 2,198 questions
- 22.2%20.6% to 24.0%, 95% interval; 489 right
- Answers kept at 99% confidence
- No answers at this thresholdnone of the 2,198 questions at 99% or more
- Cost, $ per 1,000 questionsper 1M questions
- No list priceBasis and price under Cost
- Latency
- 0.016 smedian request, 1 question per request
How we called it
- Kind
- Decision model
- Access
- GLiDE server (Fastino's release code), self-hosted
- Paid as
- Our own hardware
- Model ID
glide-0.4b- Variant
- local
- Parameters
- 400M dense, as the maker states it (counted from the released files: 0.40B).encoder (no MLM head) + 255-way readout; source: counted from the checkpoint's safetensors, fastino/GLiDE-0.4B@6bda7ab, huggingface.co/fastino/GLiDE-0.4B; read 10 November 2026
- Hardware
- 1 × DGX Spark, GB10 (128 GB unified memory); BF16
- Note
- Fastino's released GLiDE 0.4B (fastino/GLiDE-0.4B at 6bda7ab), an encoder (ModernBERT architecture, from Ettin-encoder-400m) that reads at most 7,999 tokens, run on our hardware through the server it ships.
- Confidence from
- Option score
- Run
- 931, 10 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
3b14502cefed- Prompt version
254597b8be16
Results
| Questions | Accuracy | 95% interval | Right |
|---|---|---|---|
| All questions | 22.2% | 20.6% to 24.0% | 489 of 2,198 |
| Reading | 24.5% | 21.4% to 28.0% | 161 of 656 |
| Grammar | 21.3% | 19.3% to 23.4% | 328 of 1,542 |
Comparison group: ΔΘΛΞ. Compare with another setup.
Accuracy by topic
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
- 325 of 547, first English run (933)
- Accuracy
- 59.4%
- Non-hallucination with “I don't know”
- 53.1%: 153 “I don't know” of 288 not right (run 935)
- Answered with “I don't know”
- 72.0%: 394 of 547
- Error rate with “I don't know”
- 34.3%: 135 of 394 answers
- Cost, $ per 1,000 questionsper 1M questions, first English run
- No list price
- Latency, median request, first English run
- 0.014 s; 1 request at once
- English run
- 933
- Machine, English runs
- DGX Spark
- Date
- 10 November 2026
- Request format
- 1 question per request
- Reasoning
- off
- Prompt version
47f5776e71db- Translations, SHA-256
bf0d47cd385a93050c65906977e199ef7e45d5848dbbdef2bc8d124f4d5abf68
Confidence
- Answers kept at 99% or more
- No answers at this threshold; 2,198 skipped, the model less than 99% sure
- Calibration gap
- 8.0 points
- AUROC, all questions
- 0.52 0.49 to 0.54
- AUROC, reading
- 0.52 0.47 to 0.57
- AUROC, grammar
- 0.51 0.47 to 0.54
- Brier score
- 0.184
- Accuracy of its surest answers
- most confident fifth: 23.4% right; most confident half: 22.7% right; most confident four fifths: 23.0% right
- Typical score of a right answer
- 28.2%
- Typical score of a wrong answer
- 28.1%
Repeats
We did not repeat this setup.
Latency
- Latency, median request
- 0.016 s (this run)
- Latency, 90th percentile
- 0.021 s
- Requests at once
- 1
- Throughput
- 3608.5 questions per minute of request time
- Request time
- 0.6 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
- No list price
- Basis
- No dollar estimate; see hardware and time
- Input tokens per question
- 526
- 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 | 0.0% | −0.02 points | 0%† |
| GPT-6 Astra high · chat | 0.0% | −0.02 points | 0%† |
| GPT-6.1 Sol high · chat | 0.0% | −0.02 points | 0%† |
| GPT-6 Astra low · chat | 0.0% | −0.02 points | 0%† |
| GPT-6 Astra low · chat · 1 per request | 0.0% | −0.02 points | 0%† |
| Gemini 3.8 Flash low · chat · 1 per request | 0.0% | −0.02 points | 0%† |
| GPT-6.1 Sol low · chat · 1 per request | 0.0% | −0.02 points | 0%† |
| GPT-6.1 Sol low · chat | 0.0% | −0.02 points | 0%† |
| Claude Opus 5.5 high · chat | 0.0% | −0.02 points | 0%† |
| GPT-6.1 Terra high · chat | 0.0% | −0.02 points | 0%† |
| Claude Opus 5.5 low · chat · 1 per request | 0.0% | −0.02 points | 0%† |
| Gemini 3.8 Flash low · chat | 0.0% | −0.02 points | 0%† |
| GPT-6.1 Terra low · chat | 0.0% | −0.02 points | 0%† |
| Claude Opus 5.5 low · chat | 0.0% | −0.02 points | 0%† |
| Claude Sonnet 5.5 high · chat · 1 per request | 0.1% | +0.00 points | 0%† |
| Claude Sonnet 5.5 high · chat | 0.0% | −0.03 points | 0%† |
† Leaves out the first model’s hardware cost, which we have not estimated.