Chat model
Qwen3.5-2B reasoning off · chat
reasoning off · 1 question per request
- Accuracy, all 2,198 questions
- 25.4%23.6% to 27.2%, 95% interval; 558 right
- Answers kept at 99% confidence
- Not measureda plain answer line, no confidence
- Cost, $ per 1,000 questionsper 1M questions
- $0.268$268Basis and price under Cost
- Latency
- 8.84 smedian request, 1 question per request
How we called it
- Kind
- Chat model
- Access
- our GPU (logic's server, vLLM v0.30.0 BF16, revision Qwen/Qwen3.5-2B@15852e8), one question per request, plain answer line CEVAP: X (no answer schema); sampling min_p 0.0, presence_penalty 1.5, repetition_penalty 1.0, seed 0, temperature 1.0, top_k 20, top_p 0.95, thinking off; output cap 32,768
- Paid as
- Our own hardware
- Model ID
qwen3.5-2b-nothink- Parameters
- 2B dense, as the maker states it (counted from the released files: 2.27B).language model (tied embeddings) + multi-token prediction layer + 0.33B vision encoder; source: counted from the checkpoint's safetensors, Qwen/Qwen3.5-2B@15852e8, huggingface.co/Qwen/Qwen3.5-2B; read 15 November 2026
- Hardware
- 1 × RTX PRO 6000 Blackwell (96 GB); BF16
- Confidence from
- Not measureda plain answer line, no confidence
- Run
- 1218, 15 November 2026
Settings
- Reasoning
- off
- Request format
- 1 question per request
- Requests at once
- 32
- Temperature
- not recorded
Run record: harness seed, harness commit and prompt version
- Harness seed
- 0
- Harness commit
7bc1f66c76b0- Prompt version
ba455bb00e5c
Results
| Questions | Accuracy | 95% interval | Right |
|---|---|---|---|
| All questions | 25.4% | 23.6% to 27.2% | 558 of 2,198 |
| Reading | 34.3% | 30.8% to 38.0% | 225 of 656 |
| Grammar | 21.6% | 19.6% to 23.7% | 333 of 1,542 |
Comparison group: ψΓΔΘ. Compare with another setup.
Invalid answers: 154; read from broken replies: 0; empty replies asked again: 0. Still invalid: 8 cut off at the output limit, 146 unreadable. An empty reply is asked once more; in a broken reply, any letter that is clearly readable is scored and the rest count as wrong. Invalid answers count as wrong (how).
With “Bilmiyorum” (run 1219): Invalid answers: 221; read from broken replies: 0; empty replies asked again: 0. Still invalid: 5 cut off at the output limit, 216 unreadable. (how)
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
- 368 of 547, first English run (1215)
- Accuracy
- 67.3%
- Non-hallucination with “I don't know”
- 10.3%: 18 “I don't know” of 174 not right (run 1216)
- Answered with “I don't know”
- 96.3%: 527 of 547
- Error rate with “I don't know”
- 29.6%: 156 of 527 answers
- Cost, $ per 1,000 questionsper 1M questions, first English run
- 0.280280from GPU time
- Latency, median request, first English run
- 9.27 s; 32 requests at once
- English run
- 1215
- Machine, English runs
- RTX PRO 6000
- Date
- 15 November 2026
- Request format
- 1 question per request
- Reasoning
- off
- Prompt version
ad6f4cc346df- Translations, SHA-256
bf0d47cd385a93050c65906977e199ef7e45d5848dbbdef2bc8d124f4d5abf68
With “I don't know” (run 1216): Invalid answers: 2; read from broken replies: 0; empty replies asked again: 0. Still invalid: 2 unreadable. (how)
Confidence
Not measured: this setup was asked for a plain answer line, so it states no confidence and reports no option probabilities. It has no answers kept at a confidence threshold, no calibration gap, ranking score (AUROC) or Brier score, and no typical score of a right or wrong answer.
Repeats
We did not repeat this setup.
Latency
- Latency, median request
- 8.84 s (this run)
- Latency, 90th percentile
- 20.1 s
- Requests at once
- 32
- Throughput
- 155.1 questions per minute of request time
- Request time
- 14.2 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
- Cost per 1,000 questionsper 1M questions
- $0.268$268
- Whole run
- $0.59
- Basis
- Estimated from GPU time: the run's request time x the GPUs its server held x Runpod's on-demand price for an RTX PRO 6000, $2.49 per GPU-hour (read 5 November 2026). A cloud-equivalent estimate, not measured spending.
- Price
- $2.49 per GPU-hour, 1 GPU, for the run's request time (source, 2026-11-05)
- Input tokens per question
- 266
- Output tokens per question
- 1,278