Benchmark Almanac

P3B3

Gemini 3 Flash 99.8 leads since March 2026
Not in model cards

Models hold conversations in Portuguese under three language instructions. The judge measures how closely replies follow European or Brazilian usage.

The workshop paper compares 20 models on 74 conversations with 203 turns. Higher scores indicate closer European usage under the European Portuguese instruction.

Released
March 2026
Kind
Chat and instructions in many languages
Level
Everyday
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
Portuguese
In and out
Text → Text
Interaction
Multi-turn
Items from
Written by experts
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Gemini 3 Flash #60 on the capability index
Last model added
April 2026
Scores refreshed
November 20th, 2026
20
Entries in its paper, 13 models we track

Gemini 3 Flash leads with 99.8 among models released by April 2026 Gemini 3 Flash leads with 99.8 among models released by April 2026

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. 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 April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Gemini 3 Flash pt-PT 99.8
  2. 2 Gemma 4 31B IT pt-PT 98.0
  3. 3 AMALIA-9B pt-PT 95.9
  4. 4 Gemma 4 E4B pt-PT 94.4
  5. 5 Qwen3.5-27B pt-PT 93.3
  6. 6 Gemma 3 12B pt-PT 92.9
  7. 7 Ministral 3 14B pt-PT 90.7
  8. 8 EuroLLM-22B pt-PT 88.5
  9. 9 Sabiá-4 pt-PT 87.2
  10. 10 Apertus 70B pt-PT 86.3
  11. 11 Qwen3.5-9B pt-PT 86.1
  12. 12 Gervásio-70B-ptpt pt-PT 75.2
  13. 13 Llama 3.3 70B pt-PT 67.8
  14. 14 Apertus 8B Instruct pt-PT 61.0
  15. 15 Salamandra-7B pt-PT 52.8
  16. 16 Llama 3.1-8B pt-PT 47.7
  17. 17 Gervásio-8B-ptpt pt-PT 46.6
  18. 18 Olmo 3.1 32B Instruct pt-PT 42.8
  19. 19 Qwen3-8B pt-PT 32.7
  20. 20 OLMo-3-7B pt-PT 23.3

Source: The P3B3 paper

#EntryScoreSettingDateSource
1 Gemini 3 Flash 99.8 pt-PT Dec 2025 Official leaderboard
2 Gemma 4 31B IT 98.0 pt-PT Apr 2026 Official leaderboard
3 AMALIA-9B open weights 95.9 pt-PT Official leaderboard
4 Gemma 4 E4B 94.4 pt-PT Mar 2026 Official leaderboard
5 Qwen3.5-27B 93.3 pt-PT Feb 2026 Official leaderboard
6 Gemma 3 12B 92.9 pt-PT Mar 2025 Official leaderboard
7 Ministral 3 14B 90.7 pt-PT Dec 2025 Official leaderboard
8 EuroLLM-22B open weights 88.5 pt-PT Official leaderboard
9 Sabiá-4 closed weights 87.2 pt-PT Official leaderboard
10 Apertus 70B 86.3 pt-PT Sep 2025 Official leaderboard
11 Qwen3.5-9B 86.1 pt-PT Feb 2026 Official leaderboard
12 Gervásio-70B-ptpt open weights 75.2 pt-PT Official leaderboard
13 Llama 3.3 70B 67.8 pt-PT Dec 2024 Official leaderboard
14 Apertus 8B Instruct 61.0 pt-PT Sep 2025 Official leaderboard
15 Salamandra-7B open weights 52.8 pt-PT Official leaderboard
16 Llama 3.1-8B 47.7 pt-PT Jul 2024 Official leaderboard
17 Gervásio-8B-ptpt open weights 46.6 pt-PT Official leaderboard
18 Olmo 3.1 32B Instruct 42.8 pt-PT Dec 2025 Official leaderboard
19 Qwen3-8B 32.7 pt-PT Apr 2025 Official leaderboard
20 OLMo-3-7B open weights 23.3 pt-PT Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 3 Flash 99.8 pt-PT Dec 2025 Official leaderboard
2 Gemma 4 31B IT 98.0 pt-PT Apr 2026 Official leaderboard
3 Gemma 4 E4B 94.4 pt-PT Mar 2026 Official leaderboard
4 Qwen3.5-27B 93.3 pt-PT Feb 2026 Official leaderboard
5 Gemma 3 12B 92.9 pt-PT Mar 2025 Official leaderboard
6 Ministral 3 14B 90.7 pt-PT Dec 2025 Official leaderboard
7 Apertus 70B 86.3 pt-PT Sep 2025 Official leaderboard
8 Qwen3.5-9B 86.1 pt-PT Feb 2026 Official leaderboard
9 Llama 3.3 70B 67.8 pt-PT Dec 2024 Official leaderboard
10 Apertus 8B Instruct 61.0 pt-PT Sep 2025 Official leaderboard
11 Llama 3.1-8B 47.7 pt-PT Jul 2024 Official leaderboard
12 Olmo 3.1 32B Instruct 42.8 pt-PT Dec 2025 Official leaderboard
13 Qwen3-8B 32.7 pt-PT Apr 2025 Official leaderboard

Rankings by part

We read 2 parts from the benchmark’s paper: PeroVaz and PtBrVId. Each ranks models on one part of the test. Gemini 3 Flash leads both. Arrows show how far a model moved from its overall rank among the models we track.

PeroVaz

13 models we track have a score here.

  1. 1 Gemini 3 Flash 79.7
  2. 2 Qwen3.5-27B ↑2 78.4
  3. 3 Qwen3.5-9B ↑5 76.0
  4. 4 Ministral 3 14B ↑2 74.8
  5. 5 Gemma 3 12B 72.2
  6. 6 Gemma 4 E4B ↓3 70.1
  7. 7 Gemma 4 31B IT ↓5 69.7
  8. 8 Apertus 70B ↓1 68.8
  9. 9 Llama 3.3 70B 61.6
  10. 10 Apertus 8B Instruct 59.6

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemini 3 Flash
pt-BR No-Prompt pt-PT
99.8
Gemma 4 31B IT
pt-BR No-Prompt pt-PT
98.0
Gemma 4 E4B
pt-BR No-Prompt pt-PT
94.4
Qwen3.5-27B
pt-BR No-Prompt pt-PT
93.3
Gemma 3 12B
pt-BR No-Prompt pt-PT
92.9
Ministral 3 14B
pt-BR No-Prompt pt-PT
90.7
Apertus 70B
pt-BR No-Prompt pt-PT
86.3
Qwen3.5-9B
pt-BR No-Prompt pt-PT
86.1
Llama 3.3 70B
pt-BR No-Prompt pt-PT
67.8
Apertus 8B Instruct
pt-BR pt-PT
61.0
Llama 3.1-8B
pt-BR pt-PT
47.7
Olmo 3.1 32B Instruct
pt-BR pt-PT
42.8

Source: The P3B3 paper

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3 Flash pt-PT 99.8
Gemma 4 31B IT pt-PT 98.0
Gemma 4 E4B pt-PT 94.4
Qwen3.5-27B pt-PT 93.3
Gemma 3 12B pt-PT 92.9
Ministral 3 14B pt-PT 90.7
Apertus 70B pt-PT 86.3
Qwen3.5-9B pt-PT 86.1
Llama 3.3 70B pt-PT 67.8
Gemini 3 Flash No-Prompt 63.7
Apertus 8B Instruct pt-PT 61.0
Gemma 4 31B IT No-Prompt 49.2
Llama 3.1-8B pt-PT 47.7
Qwen3.5-27B No-Prompt 44.5
Olmo 3.1 32B Instruct pt-PT 42.8
Qwen3.5-9B No-Prompt 35.1
Qwen3-8B pt-PT 32.7
Gemma 3 12B No-Prompt 23.4
Gemma 4 E4B No-Prompt 21.9
Apertus 8B Instruct No-Prompt 20.3
Ministral 3 14B No-Prompt 17.8
Olmo 3.1 32B Instruct No-Prompt 16.6
Apertus 70B No-Prompt 16.1
Llama 3.3 70B No-Prompt 15.3
Apertus 8B Instruct pt-BR 14.7
Llama 3.1-8B No-Prompt 12.6
Qwen3.5-9B pt-BR 12.2
Qwen3.5-27B pt-BR 11.9
Llama 3.1-8B pt-BR 10.6
Olmo 3.1 32B Instruct pt-BR 10.5
Qwen3-8B No-Prompt 10.4
Apertus 70B pt-BR 8.60
Qwen3-8B pt-BR 7.90
Gemma 4 31B IT pt-BR 7.30
Llama 3.3 70B pt-BR 6.70
Gemma 3 12B pt-BR 4.90
Ministral 3 14B pt-BR 4.20
Gemma 4 E4B pt-BR 4.20
Gemini 3 Flash pt-BR 4.10

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemini 3 Flash
pt-BR No-Prompt pt-PT
99.8
Gemma 4 31B IT
pt-BR No-Prompt pt-PT
98.0
Gemma 4 E4B
pt-BR No-Prompt pt-PT
94.4
Qwen3.5-27B
pt-BR No-Prompt pt-PT
93.3
Gemma 3 12B
pt-BR No-Prompt pt-PT
92.9
Ministral 3 14B
pt-BR No-Prompt pt-PT
90.7
Apertus 70B
pt-BR No-Prompt pt-PT
86.3
Qwen3.5-9B
pt-BR No-Prompt pt-PT
86.1
Llama 3.3 70B
pt-BR No-Prompt pt-PT
67.8
Apertus 8B Instruct
pt-BR pt-PT
61.0
Llama 3.1-8B
pt-BR pt-PT
47.7
Olmo 3.1 32B Instruct
pt-BR pt-PT
42.8

Source: The P3B3 paper

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3 Flash pt-PT 99.8
Gemma 4 31B IT pt-PT 98.0
Gemma 4 E4B pt-PT 94.4
Qwen3.5-27B pt-PT 93.3
Gemma 3 12B pt-PT 92.9
Ministral 3 14B pt-PT 90.7
Apertus 70B pt-PT 86.3
Qwen3.5-9B pt-PT 86.1
Llama 3.3 70B pt-PT 67.8
Gemini 3 Flash No-Prompt 63.7
Apertus 8B Instruct pt-PT 61.0
Gemma 4 31B IT No-Prompt 49.2
Llama 3.1-8B pt-PT 47.7
Qwen3.5-27B No-Prompt 44.5
Olmo 3.1 32B Instruct pt-PT 42.8
Qwen3.5-9B No-Prompt 35.1
Qwen3-8B pt-PT 32.7
Gemma 3 12B No-Prompt 23.4
Gemma 4 E4B No-Prompt 21.9
Apertus 8B Instruct No-Prompt 20.3
Ministral 3 14B No-Prompt 17.8
Olmo 3.1 32B Instruct No-Prompt 16.6
Apertus 70B No-Prompt 16.1
Llama 3.3 70B No-Prompt 15.3
Apertus 8B Instruct pt-BR 14.7
Llama 3.1-8B No-Prompt 12.6
Qwen3.5-9B pt-BR 12.2
Qwen3.5-27B pt-BR 11.9
Llama 3.1-8B pt-BR 10.6
Olmo 3.1 32B Instruct pt-BR 10.5
Qwen3-8B No-Prompt 10.4
Apertus 70B pt-BR 8.60
Qwen3-8B pt-BR 7.90
Gemma 4 31B IT pt-BR 7.30
Llama 3.3 70B pt-BR 6.70
Gemma 3 12B pt-BR 4.90
Ministral 3 14B pt-BR 4.20
Gemma 4 E4B pt-BR 4.20
Gemini 3 Flash pt-BR 4.10

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Google DeepMind
Benchmark paper Not tested since Apr 2026

Source: The P3B3 paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper
0.00 20.0 40.0 60.0 80.0 100.0 2025 2026 0.00 20.0 40.0 60.0 80.0 100.0 2025 2026

Source: Official P3B3 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

All 545 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 7 months ago, in April 2026, for Gemma 4 31B IT. 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 benchmark’s paper (20 entries).
  • Pass Models scored 13 models we track, plus 7 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for Gemma 4 31B IT.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Flash (#60).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 39 rows for models we track and 7 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.

How we check