arm
ag_qwen_l_n200_v1
One config on ag_news_200, scored one way; its repeats averaged per item.
- dataset
- ag_news_200
- scored against
- gold/ag_news_200 gold
- goal
- News topic classification (AG News)
- runs
- 1
- headline
- correct 0.870
- last run
- 2026-09-27 10:46
Inspect itAgainst the baselineWith the 33 arms scored the same way
01metrics
| metric | mean | items | better |
|---|---|---|---|
| correct | 0.870 | 200 | higher |
| answer_words | 1.000 | 200 | neutral |
| correct_after_repair | 0.000 | 200 | neutral |
| cost_usd | $0.000082 | 200 | lower |
| fmt_markdown | 0.000 | 200 | neutral |
| fmt_verbose | 0.000 | 200 | neutral |
| latency_ms | 2849.4 | 200 | lower |
| parsed | 1.000 | 200 | higher |
| reasoning_tokens | 0.000 | 200 | neutral |
| tokens_in | 98.4 | 200 | lower |
| tokens_out | 1.370 | 200 | lower |
| truncated | 0.000 | 200 | neutral |
02cost and speed
- cost per item
- $0.000082 · 100 tok
- speed per item
- p50 2,895 · p95 4,011 ms
Per item, over its items: cost as dollars and tokens (a local model is free but still uses tokens), speed as the median and the slow tail (p95). A goal that names this arm weighs them in its decision.
03runs
| run | made | items | cost | how |
|---|---|---|---|---|
ag_qwen_l_n200_v1_20260927T103653Z | 2026-09-27 10:46 | 200 | $0.016425 | measured |
04config
config ag_qwen_l_n200_v1 · adapter examples/classification-ag-news/adapter.py · credentials are never shown
- family
qwen- max_tokens
200- model
eval-qwen-3max- price_tier
large- prompt_file
prompt.txt- provider
litellm- reasoning
{"enabled": false}- temperature
0.0- usd_per_mtok_in
0.4- usd_per_mtok_out
1.6
05conditions
- data.items_sha256
d7bf61eb3b27e79d492bbe5a445659813dfcacf6fadcae4d26255c45291bedd2- deployment.model
eval-qwen-3max- deployment.source
adapter- deployment.upstream
openrouter/qwen/qwen3-max- fingerprint.version
5- library.numpy
1.26.4- library.openai
3.19.2- library.torch
2.2.2- library.transformers
4.55.4- machine.cpu_count
12- machine.label
mac-dev- machine.platform
Darwin/x86_64- machine.processor
i386- machine.python
3.11.8- machine.release
24.6.0- reference.sha256
e671b1a159664de82a0a5f35d36150d81abc99d45ac19e7df8ec1d66164bcc04- scorer.id
examples/classification-ag-news/adapter.py- scorer.sha256
8e4dffce35059b76949a0c95860cbcb5bd372fbdb6bdf33e0e55f4df78045b94
06the same config, scored other ways
scored against silverscored by a judgescored against a label set, not a reference setunmarked rows are scored against gold
| arm | scored against | headline | mean |
|---|---|---|---|
| ag_qwen_l_n200_v1 · adjudicated | gold/ag_news_200_adj1 gold | correct | 0.955 |
07rounds
- Round r1 of News topic classification (AG News), closed: one of its arms.