Benchmark Almanac

INVALSI Italian LLM Benchmark

Claude 3.5 Sonnet 92.2% leads since June 2024
Not in model cards

Models answer Italian reading and grammar questions from 11 school tests. For questions with several parts, the model must answer every part correctly.

The repository lists 405 questions and 618 items across six grades. The leaderboard compares 45 models.

Released
June 2024
Kind
Reading comprehension
Level
School
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
Italian
In and out
Text → Text
Interaction
Single turn
Items from
Exams and contests
Refresh
One fixed set
Run by
Anyone
Strongest model tested
o1-mini #142 on the capability index
Last model added
March 2025
Scores refreshed
November 20th, 2026
45
Entries on its own leaderboard, 31 models we track

Claude 3.5 Sonnet leads with 92.2% among models released by March 2025 Claude 3.5 Sonnet leads with 92.2% among models released by March 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 March 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Claude 3.5 Sonnet 92.2%
  2. 2 o1-preview 90.4%
  3. 3 Claude 3 Opus 88.5%
  4. 4 Mistral Large 2 Jul 2024 87.5%
  5. 5 Llama 3.1-405B 86.1%
  6. 6 GPT-4 Turbo Apr 2024 86%
  7. 7 Claude 3 Sonnet 83.1%
  8. 8 Llama 3.2 90B 82.9%
  9. 9 Llama 3.1-70B 82.7%
  10. 10 google/gemini-pro-1.5 81.2%
  11. 11 google/gemini-flash-1.5 80.9%
  12. 12 Mixtral 8x22B 80.9%
  13. 13 o1-mini 80.9%
  14. 14 Gemma 2 27B 80.7%
  15. 15 Qwen2-72B 79.5%
  16. 16 GPT-4o mini 79%
  17. 17 cognitivecomputations/dolphin-2.9.1-llama-3-70b 78%
  18. 18 Claude 3 Haiku 78%
  19. 19 Gemini 1.0 Pro 76.5%
  20. 20 GPT-4o Mar 2025 75.8%
  21. 21 Llama 3-70B 75.6%
  22. 22 Command R+ 75%
  23. 23 Gemma 2 9B 73.8%
  24. 24 Nemotron-4-340B-Instruct 72.1%
  25. 25 Mixtral 8x7B 68.5%
  26. 26 Llama 3.2 11B 62.4%
  27. 27 Mistral NeMo 61.9%
  28. 28 DeepMount00/Llama-3.1-8b-Ita 61.8%
  29. 29 GPT-3.5 Turbo Jan 2024 61.1%
  30. 30 DeepMount00/Llama-3-8b-Ita 59.9%
  31. 31 Llama 3.1-8B 56.5%
  32. 32 swap-uniba/LLaMAntino-3-ANITA-8B-Inst-DPO-ITA 56%
  33. 33 mii-llm/maestrale-chat-v0.4-beta 55.5%
  34. 34 mistralai/Mistral-7B-Instruct-v0.3 50.6%
  35. 35 Phi-3-medium-128k-instruct 50.6%
  36. 36 Llama 3-8B 47.9%
  37. 37 mii-community/zefiro-7b-base-ITA 44.2%
  38. 38 iGeniusAI/Italia-9B-Instruct-v0.1 37.7%
  39. 39 DeepMount00/Llama-3-COT-ITA 35.2%
  40. 40 Phi-3-mini-128k-instruct 33.5%
  41. 41 ludocomito/Minerva_3B_Ties_1.0 30.1%
  42. 42 sapienzanlp/modello-italia-9b 25.7%
  43. 43 Llama 3.2 3B 25.4%
  44. 44 Llama 3.2 1B 15.6%
  45. 45 sapienzanlp/Minerva-3B-base-v1.0 4.90%

Source: Official INVALSI Italian LLM Benchmark leaderboard

#EntryScoreSettingDateSource
1 Claude 3.5 Sonnet 92.2% Jun 2024 Official leaderboard
2 o1-preview 90.4% Sep 2024 Official leaderboard
3 Claude 3 Opus 88.5% Feb 2024 Official leaderboard
4 Mistral Large 2 Jul 2024 87.5% Jul 2024 Official leaderboard
5 Llama 3.1-405B 86.1% Jul 2024 Official leaderboard
6 GPT-4 Turbo Apr 2024 86% Apr 2024 Official leaderboard
7 Claude 3 Sonnet 83.1% Feb 2024 Official leaderboard
8 Llama 3.2 90B 82.9% Sep 2024 Official leaderboard
9 Llama 3.1-70B 82.7% Jul 2024 Official leaderboard
10 google/gemini-pro-1.5 closed weights 81.2% Official leaderboard
11 google/gemini-flash-1.5 closed weights 80.9% Official leaderboard
12 Mixtral 8x22B 80.9% Apr 2024 Official leaderboard
13 o1-mini 80.9% Sep 2024 Official leaderboard
14 Gemma 2 27B 80.7% Jun 2024 Official leaderboard
15 Qwen2-72B 79.5% Jun 2024 Official leaderboard
16 GPT-4o mini 79% Jul 2024 Official leaderboard
17 cognitivecomputations/dolphin-2.9.1-llama-3-70b open weights 78% Official leaderboard
18 Claude 3 Haiku 78% Mar 2024 Official leaderboard
19 Gemini 1.0 Pro 76.5% Dec 2023 Official leaderboard
20 GPT-4o Mar 2025 75.8% Mar 2025 Official leaderboard
21 Llama 3-70B 75.6% Apr 2024 Official leaderboard
22 Command R+ 75% Aug 2024 Official leaderboard
23 Gemma 2 9B 73.8% Jun 2024 Official leaderboard
24 Nemotron-4-340B-Instruct 72.1% Jun 2024 Official leaderboard
25 Mixtral 8x7B 68.5% Dec 2023 Official leaderboard
26 Llama 3.2 11B 62.4% Sep 2024 Official leaderboard
27 Mistral NeMo 61.9% Jul 2024 Official leaderboard
28 DeepMount00/Llama-3.1-8b-Ita open weights 61.8% Official leaderboard
29 GPT-3.5 Turbo Jan 2024 61.1% Jan 2024 Official leaderboard
30 DeepMount00/Llama-3-8b-Ita open weights 59.9% Official leaderboard
31 Llama 3.1-8B 56.5% Jul 2024 Official leaderboard
32 swap-uniba/LLaMAntino-3-ANITA-8B-Inst-DPO-ITA open weights 56% Official leaderboard
33 mii-llm/maestrale-chat-v0.4-beta open weights 55.5% Official leaderboard
34 mistralai/Mistral-7B-Instruct-v0.3 open weights 50.6% Official leaderboard
35 Phi-3-medium-128k-instruct 50.6% May 2024 Official leaderboard
36 Llama 3-8B 47.9% Apr 2024 Official leaderboard
37 mii-community/zefiro-7b-base-ITA open weights 44.2% Official leaderboard
38 iGeniusAI/Italia-9B-Instruct-v0.1 open weights 37.7% Official leaderboard
39 DeepMount00/Llama-3-COT-ITA open weights 35.2% Official leaderboard
40 Phi-3-mini-128k-instruct 33.5% Apr 2024 Official leaderboard
41 ludocomito/Minerva_3B_Ties_1.0 open weights 30.1% Official leaderboard
42 sapienzanlp/modello-italia-9b open weights 25.7% Official leaderboard
43 Llama 3.2 3B 25.4% Sep 2024 Official leaderboard
44 Llama 3.2 1B 15.6% Sep 2024 Official leaderboard
45 sapienzanlp/Minerva-3B-base-v1.0 open weights 4.90% Official leaderboard
#ModelScoreSettingDateSource
1 Claude 3.5 Sonnet 92.2% Jun 2024 Official leaderboard
2 o1-preview 90.4% Sep 2024 Official leaderboard
3 Claude 3 Opus 88.5% Feb 2024 Official leaderboard
4 Mistral Large 2 Jul 2024 87.5% Jul 2024 Official leaderboard
5 Llama 3.1-405B 86.1% Jul 2024 Official leaderboard
6 GPT-4 Turbo Apr 2024 86% Apr 2024 Official leaderboard
7 Claude 3 Sonnet 83.1% Feb 2024 Official leaderboard
8 Llama 3.2 90B 82.9% Sep 2024 Official leaderboard
9 Llama 3.1-70B 82.7% Jul 2024 Official leaderboard
10 Mixtral 8x22B 80.9% Apr 2024 Official leaderboard
11 o1-mini 80.9% Sep 2024 Official leaderboard
12 Gemma 2 27B 80.7% Jun 2024 Official leaderboard
13 Qwen2-72B 79.5% Jun 2024 Official leaderboard
14 GPT-4o mini 79% Jul 2024 Official leaderboard
15 Claude 3 Haiku 78% Mar 2024 Official leaderboard
16 Gemini 1.0 Pro 76.5% Dec 2023 Official leaderboard
17 GPT-4o Mar 2025 75.8% Mar 2025 Official leaderboard
18 Llama 3-70B 75.6% Apr 2024 Official leaderboard
19 Command R+ 75% Aug 2024 Official leaderboard
20 Gemma 2 9B 73.8% Jun 2024 Official leaderboard
21 Nemotron-4-340B-Instruct 72.1% Jun 2024 Official leaderboard
22 Mixtral 8x7B 68.5% Dec 2023 Official leaderboard
23 Llama 3.2 11B 62.4% Sep 2024 Official leaderboard
24 Mistral NeMo 61.9% Jul 2024 Official leaderboard
25 GPT-3.5 Turbo Jan 2024 61.1% Jan 2024 Official leaderboard
26 Llama 3.1-8B 56.5% Jul 2024 Official leaderboard
27 Phi-3-medium-128k-instruct 50.6% May 2024 Official leaderboard
28 Llama 3-8B 47.9% Apr 2024 Official leaderboard
29 Phi-3-mini-128k-instruct 33.5% Apr 2024 Official leaderboard
30 Llama 3.2 3B 25.4% Sep 2024 Official leaderboard
31 Llama 3.2 1B 15.6% Sep 2024 Official leaderboard

Rankings by part

We read 28 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as 2 - MC, 2 - MCC and 5 - CL. Claude 3.5 Sonnet leads 9 of them, and 9 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

2 - MC

31 models we track have a score here.

  1. 1 Claude 3 Opus ↑2 100%
  2. 2 Claude 3 Sonnet ↑5 100%
  3. 3 Claude 3 Haiku ↑12 100%
  4. 4 GPT-4 Turbo Apr 2024 ↑2 100%
  5. 5 Claude 3.5 Sonnet ↓4 100%
  6. 6 Llama 3.1-405B ↓1 100%
  7. 7 o1-preview ↓5 100%
  8. 8 Mixtral 8x7B ↑14 96.9%
  9. 9 Gemini 1.0 Pro ↑7 96.9%
  10. 10 Llama 3-70B ↑8 96.9%

2 models have held number one on the official leaderboard since its release in June 2024

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

Anthropic
Own leaderboard Passed 85% in Jun 2024, by Claude 3 Opus

Source: Official INVALSI Italian LLM Benchmark leaderboard

Claude 3 Opus on the official leaderboard passed 85% in February 2024

Models released before it came out already scored that high. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Claude 3.5 Sonnet Claude 3 Opus Gemini 1.0 Pro Mixtral 8x7B 0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Claude 3.5 Sonnet Gemini 1.0 Pro

Source: Official INVALSI Italian LLM Benchmark 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 Knowledge

All 332 in Knowledge
  • MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects. Intern-S2-397B 89.8%
  • MMLU In use 57-subject multiple-choice exam questions from elementary to professional level. Scored by accuracy. Seed1.8 92.3%
  • SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy. Qwen 3.6 Max (Preview) 73.9%
  • WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1. Tencent Hy4 preview 83.9%
  • BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly. Gemma 4 31B IT 74.4%
  • HLE-Diamond In use Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers. Claude Fable 5.2 77.9%

Data audit

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (45 entries).
  • Pass Models scored 31 models we track, plus 14 other entries on its leaderboard.
  • Note Newest score The newest came 20 months ago, in March 2025, for GPT-4o. That’s expected for a saturated test.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is o1-mini (#142).
  • Note Headroom Saturated. The best is 92.2%, past 85% since February 2024.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 31 rows for models we track and 14 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