RISE Humanities Data Benchmark

A test run is a single execution of a benchmark test using a defined model configuration.
Each run represents how a particular large language model (LLM) — such as GPT-4, Claude-3, or Gemini — performed on a given task at a specific time, with specific settings.

A test run includes:

Prompt and role definition – what the model was asked to do and from what perspective (e.g. “as a historian”).
Model configuration – provider, model version, temperature, and other generation parameters.
Results – the model’s actual response and its evaluation (scores such as F1 or accuracy).
Usage and cost data – token counts and calculated API costs.
Metadata – information like the test date, benchmark name, and person who executed it.

Together, test runs make it possible to compare models, providers, and configurations across benchmarks in a transparent and reproducible way.

Result 1 of 126

Test T1740 at 2026-09-09

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

openrouter qwen/qwen3.8-27b · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision100.0%

Recall100.0%

True positives3

False positives0

False negatives0

F1 score100.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Sonnenbichler","first_name":""},"publication":{"title":"Montaghem i, Nourhomadeh","year":"","place":"","pages":"","publisher":"","format":"","editor":""},"library_reference":{"shelfmark":"","subjects":""}}

Scoring

F1 micro86.0%

F1 macro84.2%

Micro precision85.6%

Micro recall86.5%

Instances263

True positives2,088

False positives352

False negatives327

Speed

Model time, all inputs477274.0 s

Mean per input1814.73 s

Slowest input21985.50 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens460.7K

Output tokens1.3M

Total tokens1.8M

Input cost$0.1266

Output cost$2.54

Total cost$4.11

Priced from 2026-09-08 · Input $/M $0.42 · Output $/M $3.00

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 2 of 126

Test T1710 at 2026-09-08

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

openrouter z-ai/glm-5.3-flash · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision33.3%

Recall33.3%

True positives1

False positives2

False negatives2

F1 score33.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Montaghem","first_name":"Nourhomadeh"},"publication":{"title":"","year":"","place":"","pages":"","publisher":"","format":"","editor":""},"library_reference":{"shelfmark":"","subjects":""}}

Scoring

F1 micro79.2%

F1 macro78.2%

Micro precision78.2%

Micro recall80.2%

Instances263

True positives1,937

False positives539

False negatives478

Speed

Model time, all inputs7443.4 s

Mean per input28.30 s

Slowest input941.20 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens323.8K

Output tokens66.6K

Total tokens390.5K

Input cost$0.0232

Output cost$0.0149

Total cost$0.0411

Priced from 2026-09-08 · Input $/M $0.075 · Output $/M $0.25

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 3 of 126

Test T1725 at 2026-09-08

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

openrouter qwen/qwen3.8-flash · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision20.0%

Recall33.3%

True positives1

False positives4

False negatives2

F1 score25.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Montaghemi","first_name":"Nourhomadeh"},"publication":{"title":"S","year":"","place":"","pages":"","publisher":"","format":"","editor":null},"library_reference":{"shelfmark":"","subjects":"Sonnenbichler - Montaghemi, Nourhomadeh"}}

Scoring

F1 micro85.3%

F1 macro83.3%

Micro precision84.2%

Micro recall86.4%

Instances258

True positives2,046

False positives384

False negatives321

Speed

Model time, all inputs32273.5 s

Mean per input122.71 s

Slowest input955.87 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens484.4K

Output tokens848.8K

Total tokens1.3M

Input cost$0.0262

Output cost$0.2523

Total cost$0.4649

Priced from 2026-09-08 · Input $/M $0.15 · Output $/M $0.47

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 4 of 126

Test T1695 at 2026-09-08

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

openrouter meta/muse-spark-1.3 · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision25.0%

Recall33.3%

True positives1

False positives3

False negatives2

F1 score29.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Montaghemi","first_name":"Nourhomadeh"},"publication":{"title":"Sonnenbichler - Montaghemi, Nourhomadeh","year":"","place":"","pages":"","publisher":"","format":"","editor":""},"library_reference":{"shelfmark":"","subjects":""}}

Scoring

F1 micro91.1%

F1 macro90.2%

Micro precision91.3%

Micro recall90.9%

Instances263

True positives2,196

False positives210

False negatives219

Speed

Model time, all inputs5234.2 s

Mean per input19.90 s

Slowest input47.48 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens454K

Output tokens630.8K

Total tokens1.1M

Input cost$0.5675

Output cost$2.68

Total cost$3.25

Priced from 2026-09-08 · Input $/M $1.25 · Output $/M $4.25

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 5 of 126

Test T1755 at 2026-09-08

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

deepseek deepseek-v4-flash-vision-exp · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision33.3%

Recall33.3%

True positives1

False positives2

False negatives2

F1 score33.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Montaghemi","first_name":"Nourhomadeh"},"publication":{"title":"","year":"","place":"","pages":"","publisher":"","format":"","editor":null},"library_reference":{"shelfmark":"","subjects":""}}

Scoring

F1 micro88.4%

F1 macro88.0%

Micro precision87.5%

Micro recall89.4%

Instances263

True positives2,158

False positives309

False negatives257

Speed

Model time, all inputs4906.1 s

Mean per input18.65 s

Slowest input52.93 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens429.7K

Output tokens523.6K

Total tokens953.3K

Input cost$0.1891

Output cost$0.6912

Total cost$0.8802

Priced from 2026-09-08 · Input $/M $0.44 · Output $/M $1.32

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 6 of 126

Test T1665 at 2026-09-04

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

openai gpt-6-astra · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision33.3%

Recall33.3%

True positives1

False positives2

False negatives2

F1 score33.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Montaghemi","first_name":"Nourhomadeh"},"publication":{"title":"","year":"","place":"","pages":"","publisher":"","format":"","editor":""},"library_reference":{"shelfmark":"","subjects":""}}

Scoring

F1 micro91.7%

F1 macro90.9%

Micro precision90.8%

Micro recall92.6%

Instances263

True positives2,237

False positives228

False negatives178

Speed

Model time, all inputs1565.3 s

Mean per input5.95 s

Slowest input15.14 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens444.1K

Output tokens73K

Total tokens517.2K

Input cost$4.44

Output cost$3.65

Total cost$8.09

Priced from 2026-09-03 · Input $/M $10.00 · Output $/M $50.00

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 7 of 126

Test T1650 at 2026-09-03

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

genai gemini-3.8-flash · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision25.0%

Recall33.3%

True positives1

False positives3

False negatives2

F1 score29.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Montaghemi","first_name":"Nourhomadeh"},"publication":{"title":"Sonnenbichler - Montaghemi, Nourhomadeh","year":"","place":"","pages":"","publisher":"","format":null,"editor":null},"library_reference":{"shelfmark":"","subjects":""}}

Scoring

F1 micro90.5%

F1 macro89.6%

Micro precision89.0%

Micro recall91.9%

Instances263

True positives2,220

False positives273

False negatives195

Speed

Model time, all inputs1726.5 s

Mean per input6.56 s

Slowest input113.52 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens406.2K

Output tokens48.8K

Total tokens455K

Input cost$0.3047

Output cost$0.183

Total cost$0.4876

Priced from 2026-09-02 · Input $/M $0.75 · Output $/M $3.75

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 8 of 126

Test T1620 at 2026-08-18

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

openrouter meta/muse-spark-1.2 · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision33.3%

Recall33.3%

True positives1

False positives2

False negatives2

F1 score33.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Montaghemi","first_name":"Nourhomadeh"},"publication":{"title":"","year":"","place":"","pages":"","publisher":"","format":"","editor":""},"library_reference":{"shelfmark":"","subjects":""}}

Scoring

F1 micro90.3%

F1 macro89.2%

Micro precision90.5%

Micro recall90.2%

Instances262

True positives2,170

False positives229

False negatives237

Speed

Model time, all inputs128221.0 s

Mean per input487.53 s

Slowest input1370.89 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens489.3K

Output tokens979K

Total tokens1.5M

Input cost$0.6117

Output cost$4.16

Total cost$4.77

Priced from 2026-08-18 · Input $/M $1.25 · Output $/M $4.25

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 9 of 126

Test T1590 at 2026-08-18

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

genai gemini-3.7-flash · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision25.0%

Recall33.3%

True positives1

False positives3

False negatives2

F1 score29.0%

Total fields12

{"type":{"type":"Reference"},"author":{"last_name":"Montaghemi","first_name":"Nourhomadeh"},"publication":{"title":"Sonnenbichler - Montaghemi, Nourhomadeh","year":"","place":"","pages":"","publisher":"","format":"","editor":null},"library_reference":{"shelfmark":"","subjects":""}}

Scoring

F1 micro88.4%

F1 macro87.5%

Micro precision86.7%

Micro recall90.2%

Instances263

True positives2,178

False positives333

False negatives237

Speed

Model time, all inputs910.4 s

Mean per input3.46 s

Slowest input10.15 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens406.2K

Output tokens49.6K

Total tokens455.8K

Input cost$0.3047

Output cost$0.1858

Total cost$0.4905

Priced from 2026-08-18 · Input $/M $0.75 · Output $/M $3.75

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Result 10 of 126

Test T1635 at 2026-08-18

index-card information-extraction typed, printed, handwritten 20, 19 de, fr, en, la, el, fi, sv, pl index bibliographic

openrouter z-ai/glm-5v-turbo · temp 0.0 · dataclass Document

Response vs ground truth

One model response from this run. The per-input comparison needs JavaScript.

Precision25.0%

Recall33.3%

True positives1

False positives3

False negatives2

F1 score29.0%

Total fields12

{
  "type": {
    "type": "Reference"
  },
  "author": {
    "last_name": "Montaghemi",
    "first_name": "Nourhomadeh"
  },
  "publication": {
    "title": "Sonnenbichler - Montaghemi, Nourhomadeh",
    "place": "",
    "pages": "",
    "publisher": "",
    "format": ""
  },
  "library_reference": {
    "shelfmark": "",
    "subjects": ""
  }
}

Scoring

F1 micro84.8%

F1 macro84.2%

Micro precision82.9%

Micro recall86.8%

Instances261

True positives2,079

False positives428

False negatives317

Speed

Model time, all inputs9303.7 s

Mean per input35.38 s

Slowest input66.46 s

Inputs timed263

Summed model time, not elapsed: requests run in parallel and the worker count is not recorded, so the run finished sooner than this.

Costs / Pricing

Input tokens519.6K

Output tokens324.4K

Total tokens843.9K

Input cost$0.6235

Output cost$1.30

Total cost$1.92

Priced from 2026-08-18 · Input $/M $1.20 · Output $/M $4.00

Cite: Hindermann, Marti, Alberto, et al., (2026). RISE-UNIBAS/humanities_data_benchmark, 10.5281/zenodo.21975827

Search Test Runs

Search Results
Show compact results Refine Search New Search

Download JSON Download CSV

Test T1740 at 2026-09-09

Test T1710 at 2026-09-08

Test T1725 at 2026-09-08

Test T1695 at 2026-09-08

Test T1755 at 2026-09-08

Test T1665 at 2026-09-04

Test T1650 at 2026-09-03

Test T1620 at 2026-08-18

Test T1590 at 2026-08-18

Test T1635 at 2026-08-18

Search Test Runs

Search Results Show compact results Refine Search New Search Download Download JSON Download CSV

Test T1740 at 2026-09-09

Test T1710 at 2026-09-08

Test T1725 at 2026-09-08

Test T1695 at 2026-09-08

Test T1755 at 2026-09-08

Test T1665 at 2026-09-04

Test T1650 at 2026-09-03

Test T1620 at 2026-08-18

Test T1590 at 2026-08-18

Test T1635 at 2026-08-18

Search Results
Show compact results Refine Search New Search

Download JSON Download CSV