Estonian AI Barometer
Models answer a visitor’s prompt in Estonian. The visitor compares two anonymous replies and votes before seeing which models wrote them.
The leaderboard reports 60 models and 17,422 votes. Its default table displays 55 models with older entries hidden.
- Made by
- University of Tartu, Tallinn University of Technology, Tallinn University, and Institute of the Estonian Language
- Released
- May 2025
- Kind
- Chat and instructions in many languages
- Level
- Everyday
- Scored
- Ranked by votes
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Preference
- Language
- Estonian
- Interaction
- Single turn
- Items from
- Live play or votes, Real work and use
- Refresh
- Live
- Run by
- A host
- In views
- Kept fresh
- Strongest model tested
-
Gemini 3 Pro #56 on the capability index
- Last model added
- December 2025
- Scores refreshed
- November 20th, 2026
Gemini 3 Flash leads with 1,653 among models released by December 2025 Gemini 3 Flash leads with 1,653 among models released by December 2025
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. 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 December 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
- 1,653
Gemini 3 Flash
- 1,572
Gemini 3 Pro
- 1,552 Claude-4.5 Opus (20251101)
- 1,535 GPT-4.1
- 1,530 Grok 3
- 1,516 GPT-5
- 1,512 GPT-4o
- 1,498 Claude-4 Opus (20250514)
- 1,491 Claude 3.5 Sonnet (20241022)
- 1,490 Gemini-2.0-Flash-001
- 1,486 Claude-3.7 Sonnet (20250219)
- 1,470 Claude-4.5 Sonnet (20250929)
- 1,470
Mistral Large 3
- 1,464 Claude 3 Opus (20240229)
- 1,461 Grok-3 mini
- 1,457
DeepSeek-V3
- 1,455 GPT-4 Turbo
- 1,453 Gemini-2.0-Flash-Lite-001
- 1,444
Gemini 1.5 Pro
- 1,439 Claude-4 Sonnet (20250514)
- 1,439 Kimi K2
- 1,431
Llama 4 Maverick
- 1,412
Gemma 3 27B
- 1,407
Mistral Medium 3.1
- 1,405 Claude 3.5 Haiku (20241022)
- 1,394 Llama-3.1-EstLLM-8B-Instruct-1125
- 1,387 Claude-4.5 Haiku (20251001)
- 1,384 Claude 3 Haiku (20240307)
- 1,384
Apertus 70B
- 1,376
Gemini 1.5 Flash
- 1,375 llama-estllm-prototype-0825
- 1,372
Llama 4 Scout
- 1,357 Meta-Llama-3.1-405B-Instruct
- 1,347
Qwen3-235B-A22B
- 1,331
Llama 3.3 70B Instruct
- 1,327 EuroLLM-9B-Instruct
- 1,322 Apertus-EstLLM-8B-Instruct-1125
- 1,317 Mistral Large (2411)
- 1,316
Gemma 3 12B
- 1,314 Gemini-1.5-Flash-8B-001
- 1,284
Apertus 8B Instruct
- 1,282 Meta-Llama-3.1-70B-Instruct
- 1,273 Mistral Small 3.1 (2503)
- 1,249
Qwen3-32B
- 1,242 Llammas
- 1,228
Qwen2.5-72B
- 1,222 Llama-3-70B-Instruct
- 1,162 Salamandra-7B-Instruct
- 1,157 Hermes-3-Llama-3.1-8B
- 1,155 Llama-3-8B-Instruct
- 1,139 Mistral-Nemo-Instruct-2407
- 1,134 Nous-Hermes-2-Mixtral-8x7B-DPO
- 1,114 Mixtral-8x7B-Instruct-v0.1
- 1,105 Teuken-7B-Instruct-v0.4
- 1,087 Mistral-7B-Instruct-v0.3
Bars start at 1,000.
- 1
Gemini 3 Flash 1,653
- 2
Gemini 3 Pro 1,572
- 3 Claude-4.5 Opus (20251101) 1,552
- 4 GPT-4.1 1,535
- 5 Grok 3 1,530
- 6 GPT-5 1,516
- 7 GPT-4o Nov 2024 1,512
- 8 Claude-4 Opus (20250514) 1,498
- 9 Claude 3.5 Sonnet (20241022) 1,491
- 10 Gemini-2.0-Flash-001 1,490
- 11 Claude-3.7 Sonnet (20250219) 1,486
- 12 Claude-4.5 Sonnet (20250929) 1,470
- 13
Mistral Large 3 1,470
- 14 Claude 3 Opus (20240229) 1,464
- 15 Grok-3 mini 1,461
- 16
DeepSeek-V3 Mar 2025 1,457
- 17 GPT-4 Turbo Apr 2024 1,455
- 18 Gemini-2.0-Flash-Lite-001 1,453
- 19
Gemini 1.5 Pro Sep 2024 1,444
- 20 Claude-4 Sonnet (20250514) 1,439
- 21 Kimi K2 Jul 2025 1,439
- 22
Llama 4 Maverick 1,431
- 23
Gemma 3 27B 1,412
- 24
Mistral Medium 3.1 1,407
- 25 Claude 3.5 Haiku (20241022) 1,405
- 26 Llama-3.1-EstLLM-8B-Instruct-1125 1,394
- 27 Claude-4.5 Haiku (20251001) 1,387
- 28 Claude 3 Haiku (20240307) 1,384
- 29
Apertus 70B 1,384
- 30
Gemini 1.5 Flash Sep 2024 1,376
- 31 llama-estllm-prototype-0825 1,375
- 32
Llama 4 Scout 1,372
- 33 Meta-Llama-3.1-405B-Instruct 1,357
- 34
Qwen3-235B-A22B 1,347
- 35
Llama 3.3 70B Instruct 1,331
- 36 EuroLLM-9B-Instruct 1,327
- 37 Apertus-EstLLM-8B-Instruct-1125 1,322
- 38 Mistral Large (2411) 1,317
- 39
Gemma 3 12B 1,316
- 40 Gemini-1.5-Flash-8B-001 1,314
- 41
Apertus 8B Instruct 1,284
- 42 Meta-Llama-3.1-70B-Instruct 1,282
- 43 Mistral Small 3.1 (2503) 1,273
- 44
Qwen3-32B 1,249
- 45 Llammas 1,242
- 46
Qwen2.5-72B 1,228
- 47 Llama-3-70B-Instruct 1,222
- 48 Salamandra-7B-Instruct 1,162
- 49 Hermes-3-Llama-3.1-8B 1,157
- 50 Llama-3-8B-Instruct 1,155
- 51 Mistral-Nemo-Instruct-2407 1,139
- 52 Nous-Hermes-2-Mixtral-8x7B-DPO 1,134
- 53 Mixtral-8x7B-Instruct-v0.1 1,114
- 54 Teuken-7B-Instruct-v0.4 1,105
- 55 Mistral-7B-Instruct-v0.3 1,087
Source: Official Estonian AI Barometer leaderboard
- 1,653
Gemini 3 Flash
- 1,572
Gemini 3 Pro
- 1,535 GPT-4.1
- 1,530 Grok 3
- 1,516 GPT-5
- 1,512 GPT-4o
- 1,470
Mistral Large 3
- 1,461 Grok-3 mini
- 1,457
DeepSeek-V3
- 1,455 GPT-4 Turbo
- 1,444
Gemini 1.5 Pro
- 1,439 Kimi K2
- 1,431
Llama 4 Maverick
- 1,412
Gemma 3 27B
- 1,407
Mistral Medium 3.1
- 1,384
Apertus 70B
- 1,376
Gemini 1.5 Flash
- 1,372
Llama 4 Scout
- 1,347
Qwen3-235B-A22B
- 1,331
Llama 3.3 70B Instruct
- 1,316
Gemma 3 12B
- 1,284
Apertus 8B Instruct
- 1,249
Qwen3-32B
- 1,228
Qwen2.5-72B
- 1,134 Nous-Hermes-2-Mixtral-8x7B-DPO
Bars start at 1,000.
- 1
Gemini 3 Flash 1,653
- 2
Gemini 3 Pro 1,572
- 3 GPT-4.1 1,535
- 4 Grok 3 1,530
- 5 GPT-5 1,516
- 6 GPT-4o Nov 2024 1,512
- 7
Mistral Large 3 1,470
- 8 Grok-3 mini 1,461
- 9
DeepSeek-V3 Mar 2025 1,457
- 10 GPT-4 Turbo Apr 2024 1,455
- 11
Gemini 1.5 Pro Sep 2024 1,444
- 12 Kimi K2 Jul 2025 1,439
- 13
Llama 4 Maverick 1,431
- 14
Gemma 3 27B 1,412
- 15
Mistral Medium 3.1 1,407
- 16
Apertus 70B 1,384
- 17
Gemini 1.5 Flash Sep 2024 1,376
- 18
Llama 4 Scout 1,372
- 19
Qwen3-235B-A22B 1,347
- 20
Llama 3.3 70B Instruct 1,331
- 21
Gemma 3 12B 1,316
- 22
Apertus 8B Instruct 1,284
- 23
Qwen3-32B 1,249
- 24
Qwen2.5-72B 1,228
- 25 Nous-Hermes-2-Mixtral-8x7B-DPO 1,134
Source: Official Estonian AI Barometer leaderboard
| # | Entry | Score ± | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | | 1,653 1614 to 1699 | Dec 2025 | aistudio.google.com | Official leaderboard | |
| 2 | | 1,572 1527 to 1620 | Nov 2025 | aistudio.google.com | Official leaderboard | |
| 3 | Claude-4.5 Opus (20251101) | 1,552 1507 to 1606 | anthropic.com | Official leaderboard | ||
| 4 | GPT-4.1 | 1,535 1500 to 1566 | Apr 2025 | openai.com | Official leaderboard | |
| 5 | Grok 3 | 1,530 1500 to 1564 | Apr 2025 | x.ai | Official leaderboard | |
| 6 | GPT-5 | 1,516 1481 to 1549 | Aug 2025 | openai.com | Official leaderboard | |
| 7 | GPT-4o Nov 2024 | 1,512 1486 to 1545 | Nov 2024 | openai.com | Official leaderboard | |
| 8 | Claude-4 Opus (20250514) | 1,498 ±35 | anthropic.com | Official leaderboard | ||
| 9 | Claude 3.5 Sonnet (20241022) | 1,491 1461 to 1531 | anthropic.com | Official leaderboard | ||
| 10 | Gemini-2.0-Flash-001 | 1,490 1459 to 1527 | aistudio.google.com | Official leaderboard | ||
| 11 | Claude-3.7 Sonnet (20250219) | 1,486 1458 to 1518 | anthropic.com | Official leaderboard | ||
| 12 | Claude-4.5 Sonnet (20250929) | 1,470 1424 to 1520 | anthropic.com | Official leaderboard | ||
| 13 | | 1,470 1417 to 1513 | Dec 2025 | docs.mistral.ai | Official leaderboard | |
| 14 | Claude 3 Opus (20240229) | 1,464 1436 to 1505 | anthropic.com | Official leaderboard | ||
| 15 | Grok-3 mini | 1,461 1428 to 1500 | Apr 2025 | x.ai | Official leaderboard | |
| 16 | | 1,457 1422 to 1494 | Mar 2025 | api-docs.deepseek.com | Official leaderboard | |
| 17 | GPT-4 Turbo Apr 2024 | 1,455 1423 to 1491 | Apr 2024 | platform.openai.com | Official leaderboard | |
| 18 | Gemini-2.0-Flash-Lite-001 | 1,453 ±29 | aistudio.google.com | Official leaderboard | ||
| 19 | | 1,444 1404 to 1493 | Sep 2024 | aistudio.google.com | Official leaderboard | |
| 20 | Claude-4 Sonnet (20250514) | 1,439 1396 to 1478 | anthropic.com | Official leaderboard | ||
| 21 | Kimi K2 Jul 2025 | 1,439 1411 to 1479 | Jul 2025 | huggingface.co | Official leaderboard | |
| 22 | | 1,431 ±32.5 | Apr 2025 | huggingface.co | Official leaderboard | |
| 23 | | 1,412 1379 to 1460 | Mar 2025 | aistudio.google.com | Official leaderboard | |
| 24 | | 1,407 1352 to 1452 | Aug 2025 | docs.mistral.ai | Official leaderboard | |
| 25 | Claude 3.5 Haiku (20241022) | 1,405 1370 to 1449 | anthropic.com | Official leaderboard | ||
| 26 | Llama-3.1-EstLLM-8B-Instruct-1125 | 1,394 1352 to 1441 | huggingface.co | Official leaderboard | ||
| 27 | Claude-4.5 Haiku (20251001) | 1,387 1347 to 1431 | anthropic.com | Official leaderboard | ||
| 28 | Claude 3 Haiku (20240307) | 1,384 1356 to 1424 | anthropic.com | Official leaderboard | ||
| 29 | | 1,384 1315 to 1435 | Sep 2025 | huggingface.co | Official leaderboard | |
| 30 | | 1,376 1347 to 1412 | Aug 2024 | aistudio.google.com | Official leaderboard | |
| 31 | llama-estllm-prototype-0825 | 1,375 1336 to 1416 | huggingface.co | Official leaderboard | ||
| 32 | | 1,372 1335 to 1421 | Apr 2025 | huggingface.co | Official leaderboard | |
| 33 | Meta-Llama-3.1-405B-Instruct | 1,357 1326 to 1394 | ai.meta.com | Official leaderboard | ||
| 34 | | 1,347 1313 to 1392 | Apr 2025 | qwenlm.github.io | Official leaderboard | |
| 35 | | 1,331 1299 to 1371 | Dec 2024 | huggingface.co | Official leaderboard | |
| 36 | EuroLLM-9B-Instruct | 1,327 1291 to 1367 | huggingface.co | Official leaderboard | ||
| 37 | Apertus-EstLLM-8B-Instruct-1125 | 1,322 1273 to 1360 | huggingface.co | Official leaderboard | ||
| 38 | Mistral Large (2411) | 1,317 1289 to 1353 | huggingface.co | Official leaderboard | ||
| 39 | | 1,316 1235 to 1374 | Mar 2025 | aistudio.google.com | Official leaderboard | |
| 40 | Gemini-1.5-Flash-8B-001 | 1,314 1285 to 1352 | aistudio.google.com | Official leaderboard | ||
| 41 | | 1,284 1242 to 1334 | Sep 2025 | huggingface.co | Official leaderboard | |
| 42 | Meta-Llama-3.1-70B-Instruct | 1,282 1240 to 1339 | ai.meta.com | Official leaderboard | ||
| 43 | Mistral Small 3.1 (2503) | 1,273 1240 to 1312 | mistral.ai | Official leaderboard | ||
| 44 | | 1,249 1215 to 1287 | Apr 2025 | qwenlm.github.io | Official leaderboard | |
| 45 | Llammas | 1,242 ±37.5 | huggingface.co | Official leaderboard | ||
| 46 | | 1,228 1189 to 1272 | Sep 2024 | qwenlm.github.io | Official leaderboard | |
| 47 | Llama-3-70B-Instruct | 1,222 1192 to 1257 | llama.meta.com | Official leaderboard | ||
| 48 | Salamandra-7B-Instruct | 1,162 1121 to 1215 | huggingface.co | Official leaderboard | ||
| 49 | Hermes-3-Llama-3.1-8B | 1,157 1119 to 1192 | huggingface.co | Official leaderboard | ||
| 50 | Llama-3-8B-Instruct | 1,155 1121 to 1195 | llama.meta.com | Official leaderboard | ||
| 51 | Mistral-Nemo-Instruct-2407 | 1,139 1089 to 1196 | huggingface.co | Official leaderboard | ||
| 52 | Nous-Hermes-2-Mixtral-8x7B-DPO | 1,134 1101 to 1181 | Jan 2024 | huggingface.co | Official leaderboard | |
| 53 | Mixtral-8x7B-Instruct-v0.1 | 1,114 | mistral.ai | Official leaderboard | ||
| 54 | Teuken-7B-Instruct-v0.4 | 1,105 ±45.5 | huggingface.co | Official leaderboard | ||
| 55 | Mistral-7B-Instruct-v0.3 | 1,087 1058 to 1127 | huggingface.co | Official leaderboard |
| # | Model | Score ± | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | | 1,653 1614 to 1699 | Dec 2025 | aistudio.google.com | Official leaderboard | |
| 2 | | 1,572 1527 to 1620 | Nov 2025 | aistudio.google.com | Official leaderboard | |
| 3 | GPT-4.1 | 1,535 1500 to 1566 | Apr 2025 | openai.com | Official leaderboard | |
| 4 | Grok 3 | 1,530 1500 to 1564 | Apr 2025 | x.ai | Official leaderboard | |
| 5 | GPT-5 | 1,516 1481 to 1549 | Aug 2025 | openai.com | Official leaderboard | |
| 6 | GPT-4o Nov 2024 | 1,512 1486 to 1545 | Nov 2024 | openai.com | Official leaderboard | |
| 7 | | 1,470 1417 to 1513 | Dec 2025 | docs.mistral.ai | Official leaderboard | |
| 8 | Grok-3 mini | 1,461 1428 to 1500 | Apr 2025 | x.ai | Official leaderboard | |
| 9 | | 1,457 1422 to 1494 | Mar 2025 | api-docs.deepseek.com | Official leaderboard | |
| 10 | GPT-4 Turbo Apr 2024 | 1,455 1423 to 1491 | Apr 2024 | platform.openai.com | Official leaderboard | |
| 11 | | 1,444 1404 to 1493 | Sep 2024 | aistudio.google.com | Official leaderboard | |
| 12 | Kimi K2 Jul 2025 | 1,439 1411 to 1479 | Jul 2025 | huggingface.co | Official leaderboard | |
| 13 | | 1,431 ±32.5 | Apr 2025 | huggingface.co | Official leaderboard | |
| 14 | | 1,412 1379 to 1460 | Mar 2025 | aistudio.google.com | Official leaderboard | |
| 15 | | 1,407 1352 to 1452 | Aug 2025 | docs.mistral.ai | Official leaderboard | |
| 16 | | 1,384 1315 to 1435 | Sep 2025 | huggingface.co | Official leaderboard | |
| 17 | | 1,376 1347 to 1412 | Aug 2024 | aistudio.google.com | Official leaderboard | |
| 18 | | 1,372 1335 to 1421 | Apr 2025 | huggingface.co | Official leaderboard | |
| 19 | | 1,347 1313 to 1392 | Apr 2025 | qwenlm.github.io | Official leaderboard | |
| 20 | | 1,331 1299 to 1371 | Dec 2024 | huggingface.co | Official leaderboard | |
| 21 | | 1,316 1235 to 1374 | Mar 2025 | aistudio.google.com | Official leaderboard | |
| 22 | | 1,284 1242 to 1334 | Sep 2025 | huggingface.co | Official leaderboard | |
| 23 | | 1,249 1215 to 1287 | Apr 2025 | qwenlm.github.io | Official leaderboard | |
| 24 | | 1,228 1189 to 1272 | Sep 2024 | qwenlm.github.io | Official leaderboard | |
| 25 | Nous-Hermes-2-Mixtral-8x7B-DPO | 1,134 1101 to 1181 | Jan 2024 | huggingface.co | Official leaderboard |
3 models have held number one on the official leaderboard since its release in May 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Estonian AI Barometer 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 Estonian AI Barometer 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 11 months ago, in December 2025, for Gemini 3 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 (55 entries).
- Pass Models scored 25 models we track, plus 30 other entries on its leaderboard.
- Warning Newest score The newest came 11 months ago, in December 2025, for Gemini 3 Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Pro (#56).
- Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on baromeeter.tartunlp.ai: 25 rows for models we track and 30 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.