ParlaSent
Models classify the sentiment of parliamentary sentences as positive, neutral, or negative. Two additional test sets contain 2,600 English and 2,600 South Slavic sentences.
The score file covers 15 language model entries plus five classifiers or baselines. It reports micro and macro F1 by language.
- Made by
- Michal Mochtak and the ParlaSent team
- Released
- September 2023
- Kind
- Reading in many languages
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Language
- Many languages
- Interaction
- Single turn
- Items from
- Real work and use, Written by crowdworkers
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Many languages
- Strongest model tested
- GPT-5 #71 on the capability index
- Last model added
- September 2025
- Scores refreshed
- November 20th, 2026
GPT-5 leads with 0.77 among models released by September 2025 GPT-5 leads with 0.77 among models released by September 2025
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The authors give no overall score, so this is the plain mean of macro F1 on the English and BCS test sets. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 0.77 GPT-5
- 0.76 GPT-5 mini
- 0.76 GPT-4o
- 0.76
Gemini 2.5 Flash
- 0.74
Gemma 3 27B
- 0.74
Mistral Medium 3.1
- 0.73 GPT-5 nano
- 0.73
Llama 3.3 70B Instruct
- 0.72
Qwen3-32B
- 0.72 XLM-R-ParlaSent
- 0.71 GPT-4o mini
- 0.68
Llama 4 Scout
- 0.68 GPT-3.5 Turbo
- 0.61
DeepSeek-R1-Distill-Qwen-14B
- 0.52 GaMS-27B
- 0.52 GaMS-27B-quantized
- 0.35 SVC
- 0.35 COMPLEMENTNB
- 1 GPT-5 0.77
- 2 GPT-5 mini 0.76
- 3 GPT-4o Aug 2024 0.76
- 4
Gemini 2.5 Flash Sep 2025 0.76
- 5
Gemma 3 27B 0.74
- 6
Mistral Medium 3.1 0.74
- 7 GPT-5 nano 0.73
- 8
Llama 3.3 70B Instruct 0.73
- 9
Qwen3-32B 0.72
- 10 XLM-R-ParlaSent 0.72
- 11 GPT-4o mini 0.71
- 12
Llama 4 Scout 0.68
- 13 GPT-3.5 Turbo Jan 2024 0.68
- 14
DeepSeek-R1-Distill-Qwen-14B 0.61
- 15 GaMS-27B 0.52
- 16 GaMS-27B-quantized 0.52
- 17 SVC 0.35
- 18 COMPLEMENTNB 0.35
Source: Official ParlaSent leaderboard
- 1 GPT-5 0.77
- 2 GPT-5 mini 0.76
- 3 GPT-4o Aug 2024 0.76
- 4
Gemini 2.5 Flash Sep 2025 0.76
- 5
Gemma 3 27B 0.74
- 6
Mistral Medium 3.1 0.74
- 7 GPT-5 nano 0.73
- 8
Llama 3.3 70B Instruct 0.73
- 9
Qwen3-32B 0.72
- 10 GPT-4o mini 0.71
- 11
Llama 4 Scout 0.68
- 12 GPT-3.5 Turbo Jan 2024 0.68
- 13
DeepSeek-R1-Distill-Qwen-14B 0.61
Source: Official ParlaSent leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5 | 0.77 | Aug 2025 | Official leaderboard | |
| 2 | GPT-5 mini | 0.76 | Aug 2025 | Official leaderboard | |
| 3 | GPT-4o Aug 2024 | 0.76 | Aug 2024 | Official leaderboard | |
| 4 | | 0.76 | Sep 2025 | Official leaderboard | |
| 5 | | 0.74 | Mar 2025 | Official leaderboard | |
| 6 | | 0.74 | Aug 2025 | Official leaderboard | |
| 7 | GPT-5 nano | 0.73 | Aug 2025 | Official leaderboard | |
| 8 | | 0.73 | Dec 2024 | Official leaderboard | |
| 9 | | 0.72 | Apr 2025 | Official leaderboard | |
| 10 | XLM-R-ParlaSent | 0.72 | Official leaderboard | ||
| 11 | GPT-4o mini | 0.71 | Jul 2024 | Official leaderboard | |
| 12 | | 0.68 | Apr 2025 | Official leaderboard | |
| 13 | GPT-3.5 Turbo Jan 2024 | 0.68 | Jan 2024 | Official leaderboard | |
| 14 | | 0.61 | Jan 2025 | Official leaderboard | |
| 15 | GaMS-27B | 0.52 | Official leaderboard | ||
| 16 | GaMS-27B-quantized | 0.52 | Official leaderboard | ||
| 17 | SVC | 0.35 | Official leaderboard | ||
| 18 | COMPLEMENTNB | 0.35 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5 | 0.77 | Aug 2025 | Official leaderboard | |
| 2 | GPT-5 mini | 0.76 | Aug 2025 | Official leaderboard | |
| 3 | GPT-4o Aug 2024 | 0.76 | Aug 2024 | Official leaderboard | |
| 4 | | 0.76 | Sep 2025 | Official leaderboard | |
| 5 | | 0.74 | Mar 2025 | Official leaderboard | |
| 6 | | 0.74 | Aug 2025 | Official leaderboard | |
| 7 | GPT-5 nano | 0.73 | Aug 2025 | Official leaderboard | |
| 8 | | 0.73 | Dec 2024 | Official leaderboard | |
| 9 | | 0.72 | Apr 2025 | Official leaderboard | |
| 10 | GPT-4o mini | 0.71 | Jul 2024 | Official leaderboard | |
| 11 | | 0.68 | Apr 2025 | Official leaderboard | |
| 12 | GPT-3.5 Turbo Jan 2024 | 0.68 | Jan 2024 | Official leaderboard | |
| 13 | | 0.61 | Jan 2025 | Official leaderboard |
Rankings by part
We read 10 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as English: Macro F1, English: Micro F1 and Serbian: Macro F1. GPT-5 leads 8 of them, and Gemini 2.5 Flash leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.
English: Macro F1
13 models we track have a score here.
- 1 GPT-5 0.78
- 2 GPT-5 mini 0.77
- 3 GPT-4o Aug 2024 0.77
- 4
Gemma 3 27B ↑1 0.77
- 5
Gemini 2.5 Flash Sep 2025 ↓1 0.77
- 6
Mistral Medium 3.1 0.76
- 7 GPT-5 nano 0.75
- 8
Llama 3.3 70B Instruct 0.75
- 9
Qwen3-32B 0.74
- 10 GPT-4o mini 0.73
English: Micro F1
13 models we track have a score here.
- 1 GPT-5 0.78
- 2 GPT-4o Aug 2024 ↑1 0.77
- 3 GPT-5 mini ↓1 0.77
- 4
Gemini 2.5 Flash Sep 2025 0.77
- 5
Gemma 3 27B 0.77
- 6
Mistral Medium 3.1 0.76
- 7 GPT-5 nano 0.75
- 8
Llama 3.3 70B Instruct 0.74
- 9
Qwen3-32B 0.74
- 10 GPT-4o mini 0.73
Serbian: Macro F1
13 models we track have a score here.
- 1 GPT-5 0.77
- 2 GPT-5 mini 0.76
- 3 GPT-4o Aug 2024 0.74
- 4
Gemini 2.5 Flash Sep 2025 0.73
- 5 GPT-5 nano ↑2 0.72
- 6
Gemma 3 27B ↓1 0.71
- 7
Qwen3-32B ↑2 0.71
- 8
Mistral Medium 3.1 ↓2 0.70
- 9
Llama 3.3 70B Instruct ↓1 0.69
- 10 GPT-4o mini 0.68
Serbian: Micro F1
13 models we track have a score here.
- 1 GPT-5 0.78
- 2 GPT-5 mini 0.77
- 3 GPT-4o Aug 2024 0.76
- 4
Gemini 2.5 Flash Sep 2025 0.76
- 5 GPT-5 nano ↑2 0.73
- 6
Gemma 3 27B ↓1 0.73
- 7
Qwen3-32B ↑2 0.72
- 8
Mistral Medium 3.1 ↓2 0.72
- 9
Llama 3.3 70B Instruct ↓1 0.71
- 10 GPT-4o mini 0.71
Bosnian: Macro F1
13 models we track have a score here.
- 1 GPT-5 0.76
- 2 GPT-5 mini 0.73
- 3
Gemma 3 27B ↑2 0.72
- 4 GPT-4o Aug 2024 ↓1 0.70
- 5
Gemini 2.5 Flash Sep 2025 ↓1 0.69
- 6 GPT-5 nano ↑1 0.68
- 7
Llama 3.3 70B Instruct ↑1 0.67
- 8
Mistral Medium 3.1 ↓2 0.64
- 9
Llama 4 Scout ↑2 0.63
- 10
Qwen3-32B ↓1 0.61
Bosnian: Micro F1
13 models we track have a score here.
- 1 GPT-5 0.78
- 2 GPT-5 mini 0.75
- 3 GPT-4o Aug 2024 0.75
- 4
Gemma 3 27B ↑1 0.74
- 5
Gemini 2.5 Flash Sep 2025 ↓1 0.73
- 6
Llama 3.3 70B Instruct ↑2 0.71
- 7 GPT-5 nano 0.69
- 8
Mistral Medium 3.1 ↓2 0.67
- 9
Llama 4 Scout ↑2 0.66
- 10
Qwen3-32B ↓1 0.66
Croatian: Macro F1
13 models we track have a score here.
- 1
Gemini 2.5 Flash Sep 2025 ↑3 0.76
- 2 GPT-4o Aug 2024 ↑1 0.76
- 3 GPT-5 ↓2 0.76
- 4 GPT-5 mini ↓2 0.75
- 5
Mistral Medium 3.1 ↑1 0.73
- 6
Gemma 3 27B ↓1 0.73
- 7
Llama 3.3 70B Instruct ↑1 0.73
- 8 GPT-5 nano ↓1 0.71
- 9
Qwen3-32B 0.70
- 10 GPT-4o mini 0.69
Croatian: Micro F1
13 models we track have a score here.
- 1
Gemini 2.5 Flash Sep 2025 ↑3 0.79
- 2 GPT-4o Aug 2024 ↑1 0.78
- 3 GPT-5 ↓2 0.77
- 4 GPT-5 mini ↓2 0.76
- 5
Mistral Medium 3.1 ↑1 0.76
- 6
Llama 3.3 70B Instruct ↑2 0.74
- 7
Gemma 3 27B ↓2 0.74
- 8
Qwen3-32B ↑1 0.73
- 9 GPT-5 nano ↓2 0.72
- 10 GPT-4o mini 0.72
ParlaSent-BCS-test: Macro F1
13 models we track have a score here.
- 1 GPT-5 0.76
- 2 GPT-5 mini 0.75
- 3 GPT-4o Aug 2024 0.75
- 4
Gemini 2.5 Flash Sep 2025 0.75
- 5
Gemma 3 27B 0.72
- 6
Mistral Medium 3.1 0.72
- 7 GPT-5 nano 0.71
- 8
Llama 3.3 70B Instruct 0.71
- 9
Qwen3-32B 0.70
- 10 GPT-4o mini 0.68
ParlaSent-BCS-test: Micro F1
13 models we track have a score here.
- 1 GPT-5 0.78
- 2 GPT-4o Aug 2024 ↑1 0.77
- 3
Gemini 2.5 Flash Sep 2025 ↑1 0.77
- 4 GPT-5 mini ↓2 0.77
- 5
Gemma 3 27B 0.73
- 6
Mistral Medium 3.1 0.73
- 7
Llama 3.3 70B Instruct ↑1 0.73
- 8
Qwen3-32B ↑1 0.72
- 9 GPT-5 nano ↓2 0.72
- 10 GPT-4o mini 0.71
4 models have held number one on the official leaderboard since its release in September 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official ParlaSent leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official ParlaSent leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Multilingual
- MMMLU In use MMLU professionally translated into 14 languages. Multiple-choice accuracy. Claude Mythos Preview 92.7%
- WMT24++ In use English-to-55-language translation of WMT24 test data. Scored with automatic metrics such as MetricX or xCOMET.
K-EXAONE 90.5%
- MMLU-ProX In use MMLU-Pro translated into up to 29 languages with expert review. Multiple-choice accuracy per language.
Qwen3.7-Max 87%
- CLIcK In use Models answer 1,995 multiple-choice questions on Korean language and culture. Questions come from official Korean exams and textbooks in eleven categories.
Solar Open 2 90.7%
- Global MMLU Lite 3.0 In use Models answer 400 human-translated MMLU questions per language in 23 languages. Version 3.0 adds Czech, Hungarian, Oriya, Slovak and Tajik and updates Italian. Accuracy. Claude Opus 5 95.7%
- MMMB In use Models answer multiple-choice questions about images in six languages across 15 categories. The set has about 12,000 questions. LFM2.5-VL-3B 83%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (18 entries).
- Pass Models scored 13 models we track, plus 5 other entries on its leaderboard.
- Warning Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5 (#71).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 13 rows for models we track and 5 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.