Hebrew LLM Chat Leaderboard
Models summarize documents, translate into Hebrew, answer trivia, resolve pronouns, and restore vowel marks. The suite reports each task separately.
The results archive contains five scores for each of 34 models. GPT-4o judges summaries and translations. The other tasks use automatic checks.
- Made by
- Dicta and the Israeli National Program for NLP
- Released
- November 2025
- Kind
- Chat and instructions in many languages
- Level
- School
- Scored
- Graded by a model
- Answers
- Open-ended
- Measures
- Capability
- Subject
- Languages and literature
- Interaction
- Single turn
- Items from
- Another benchmark
- Refresh
- One fixed set
- Run by
- A host
- Strongest model tested
- GLM-5.3-Flash #53 on the capability index
- Last model added
- August 2026
- Scores refreshed
- November 20th, 2026
Gemini 3 Pro leads with 74.7% among models released by August 2026 Gemini 3 Pro leads with 74.7% among models released by August 2026
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 August 2026, 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
- 74.7
Gemini 3 Pro
- 74.0
Gemini 3 Flash
- 72.7 GPT-5.1
- 71.9
Gemini 2.5 Pro
- 65.5
Qwen3.5 397B-A17B
- 64.8
Gemini 2.5 Flash
- 64.2 GPT-5 mini
- 62.4 dicta-il/DictaLM-3.0-24B-Thinking
- 61.9 Claude Sonnet 4.5
- 61.5
Qwen3.5-122B-A10B
- 61.4 GPT-4o
- 61.3 GLM-5.3-Flash
- 59.9
Gemma 4 31B IT
- 58.5 GLM-5.1
- 56.4
DeepSeek-V3.1
- 56.3
Gemini 2.5 Flash-Lite
- 54.2 GLM-4.6
- 53.6
MiniMax-M2.7
- 52.8
Gemma 4 26B A4B
- 51.4
Gemma 3 27B
- 51.2 Jamba 1.7 Large
- 50.4
MiniMax-M2.5
- 49.9
Cohere Command A
- 49.9 Claude Haiku 4.5
- 48.4 gpt-oss-120b
- 46.6
Nemotron 3 Super
- 45.7 Kimi K2
- 45.3
Aya Expanse 32B
- 43.7 GPT-4o mini
- 41.8 GPT-5 nano
- 41.0
Llama 3.3 70B Instruct
- 31.3
Nemotron 3 Nano 30B-A3B
- 30.8 Jamba 1.7 Mini
- 29.9
Qwen3-30B-A3B-Thinking
- 1
Gemini 3 Pro 74.7%
- 2
Gemini 3 Flash 74%
- 3 GPT-5.1 72.7%
- 4
Gemini 2.5 Pro Jun 2025 71.9%
- 5
Qwen3.5 397B-A17B 65.5%
- 6
Gemini 2.5 Flash Sep 2025 64.8%
- 7 GPT-5 mini 64.2%
- 8 dicta-il/DictaLM-3.0-24B-Thinking 62.4%
- 9 Claude Sonnet 4.5 61.9%
- 10
Qwen3.5-122B-A10B 61.5%
- 11 GPT-4o Mar 2025 61.4%
- 12 GLM-5.3-Flash 61.3%
- 13
Gemma 4 31B IT 59.9%
- 14 GLM-5.1 58.5%
- 15
DeepSeek-V3.1 56.4%
- 16
Gemini 2.5 Flash-Lite Sep 2025 56.3%
- 17 GLM-4.6 54.2%
- 18
MiniMax-M2.7 53.6%
- 19
Gemma 4 26B A4B 52.8%
- 20
Gemma 3 27B 51.4%
- 21 Jamba 1.7 Large 51.2%
- 22
MiniMax-M2.5 50.4%
- 23
Cohere Command A 49.9%
- 24 Claude Haiku 4.5 49.9%
- 25 gpt-oss-120b 48.4%
- 26
Nemotron 3 Super 46.6%
- 27 Kimi K2 Sep 2025 45.7%
- 28
Aya Expanse 32B 45.3%
- 29 GPT-4o mini 43.7%
- 30 GPT-5 nano 41.8%
- 31
Llama 3.3 70B Instruct 41%
- 32
Nemotron 3 Nano 30B-A3B 31.3%
- 33 Jamba 1.7 Mini 30.8%
- 34
Qwen3-30B-A3B-Thinking Jul 2025 29.9%
Source: Official Hebrew LLM Chat Leaderboard leaderboard
- 74.7
Gemini 3 Pro
- 74.0
Gemini 3 Flash
- 72.7 GPT-5.1
- 71.9
Gemini 2.5 Pro
- 65.5
Qwen3.5 397B-A17B
- 64.8
Gemini 2.5 Flash
- 64.2 GPT-5 mini
- 61.9 Claude Sonnet 4.5
- 61.5
Qwen3.5-122B-A10B
- 61.4 GPT-4o
- 61.3 GLM-5.3-Flash
- 59.9
Gemma 4 31B IT
- 58.5 GLM-5.1
- 56.4
DeepSeek-V3.1
- 56.3
Gemini 2.5 Flash-Lite
- 54.2 GLM-4.6
- 53.6
MiniMax-M2.7
- 52.8
Gemma 4 26B A4B
- 51.4
Gemma 3 27B
- 51.2 Jamba 1.7 Large
- 50.4
MiniMax-M2.5
- 49.9
Cohere Command A
- 49.9 Claude Haiku 4.5
- 48.4 gpt-oss-120b
- 46.6
Nemotron 3 Super
- 45.7 Kimi K2
- 45.3
Aya Expanse 32B
- 43.7 GPT-4o mini
- 41.8 GPT-5 nano
- 41.0
Llama 3.3 70B Instruct
- 31.3
Nemotron 3 Nano 30B-A3B
- 30.8 Jamba 1.7 Mini
- 29.9
Qwen3-30B-A3B-Thinking
- 1
Gemini 3 Pro 74.7%
- 2
Gemini 3 Flash 74%
- 3 GPT-5.1 72.7%
- 4
Gemini 2.5 Pro Jun 2025 71.9%
- 5
Qwen3.5 397B-A17B 65.5%
- 6
Gemini 2.5 Flash Sep 2025 64.8%
- 7 GPT-5 mini 64.2%
- 8 Claude Sonnet 4.5 61.9%
- 9
Qwen3.5-122B-A10B 61.5%
- 10 GPT-4o Mar 2025 61.4%
- 11 GLM-5.3-Flash 61.3%
- 12
Gemma 4 31B IT 59.9%
- 13 GLM-5.1 58.5%
- 14
DeepSeek-V3.1 56.4%
- 15
Gemini 2.5 Flash-Lite Sep 2025 56.3%
- 16 GLM-4.6 54.2%
- 17
MiniMax-M2.7 53.6%
- 18
Gemma 4 26B A4B 52.8%
- 19
Gemma 3 27B 51.4%
- 20 Jamba 1.7 Large 51.2%
- 21
MiniMax-M2.5 50.4%
- 22
Cohere Command A 49.9%
- 23 Claude Haiku 4.5 49.9%
- 24 gpt-oss-120b 48.4%
- 25
Nemotron 3 Super 46.6%
- 26 Kimi K2 Sep 2025 45.7%
- 27
Aya Expanse 32B 45.3%
- 28 GPT-4o mini 43.7%
- 29 GPT-5 nano 41.8%
- 30
Llama 3.3 70B Instruct 41%
- 31
Nemotron 3 Nano 30B-A3B 31.3%
- 32 Jamba 1.7 Mini 30.8%
- 33
Qwen3-30B-A3B-Thinking Jul 2025 29.9%
Source: Official Hebrew LLM Chat Leaderboard leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 74.7% | Nov 2025 | Official leaderboard | |
| 2 | | 74% | Dec 2025 | Official leaderboard | |
| 3 | GPT-5.1 | 72.7% | Nov 2025 | Official leaderboard | |
| 4 | | 71.9% | Jun 2025 | Official leaderboard | |
| 5 | | 65.5% | Feb 2026 | Official leaderboard | |
| 6 | | 64.8% | Sep 2025 | Official leaderboard | |
| 7 | GPT-5 mini | 64.2% | Aug 2025 | Official leaderboard | |
| 8 | dicta-il/DictaLM-3.0-24B-Thinking | 62.4% | Official leaderboard | ||
| 9 | Claude Sonnet 4.5 | 61.9% | Sep 2025 | Official leaderboard | |
| 10 | | 61.5% | Feb 2026 | Official leaderboard | |
| 11 | GPT-4o Mar 2025 | 61.4% | Mar 2025 | Official leaderboard | |
| 12 | GLM-5.3-Flash | 61.3% | Aug 2026 | Official leaderboard | |
| 13 | | 59.9% | Apr 2026 | Official leaderboard | |
| 14 | GLM-5.1 | 58.5% | Apr 2026 | Official leaderboard | |
| 15 | | 56.4% | Aug 2025 | Official leaderboard | |
| 16 | | 56.3% | Sep 2025 | Official leaderboard | |
| 17 | GLM-4.6 | 54.2% | Sep 2025 | Official leaderboard | |
| 18 | | 53.6% | Mar 2026 | Official leaderboard | |
| 19 | | 52.8% | Apr 2026 | Official leaderboard | |
| 20 | | 51.4% | Mar 2025 | Official leaderboard | |
| 21 | Jamba 1.7 Large | 51.2% | Jul 2025 | Official leaderboard | |
| 22 | | 50.4% | Feb 2026 | Official leaderboard | |
| 23 | | 49.9% | Mar 2025 | Official leaderboard | |
| 24 | Claude Haiku 4.5 | 49.9% | Oct 2025 | Official leaderboard | |
| 25 | gpt-oss-120b | 48.4% | Aug 2025 | Official leaderboard | |
| 26 | | 46.6% | Mar 2026 | Official leaderboard | |
| 27 | Kimi K2 Sep 2025 | 45.7% | Sep 2025 | Official leaderboard | |
| 28 | | 45.3% | Oct 2024 | Official leaderboard | |
| 29 | GPT-4o mini | 43.7% | Jul 2024 | Official leaderboard | |
| 30 | GPT-5 nano | 41.8% | Aug 2025 | Official leaderboard | |
| 31 | | 41% | Dec 2024 | Official leaderboard | |
| 32 | | 31.3% | Dec 2025 | Official leaderboard | |
| 33 | Jamba 1.7 Mini | 30.8% | Jul 2025 | Official leaderboard | |
| 34 | | 29.9% | Jul 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 74.7% | Nov 2025 | Official leaderboard | |
| 2 | | 74% | Dec 2025 | Official leaderboard | |
| 3 | GPT-5.1 | 72.7% | Nov 2025 | Official leaderboard | |
| 4 | | 71.9% | Jun 2025 | Official leaderboard | |
| 5 | | 65.5% | Feb 2026 | Official leaderboard | |
| 6 | | 64.8% | Sep 2025 | Official leaderboard | |
| 7 | GPT-5 mini | 64.2% | Aug 2025 | Official leaderboard | |
| 8 | Claude Sonnet 4.5 | 61.9% | Sep 2025 | Official leaderboard | |
| 9 | | 61.5% | Feb 2026 | Official leaderboard | |
| 10 | GPT-4o Mar 2025 | 61.4% | Mar 2025 | Official leaderboard | |
| 11 | GLM-5.3-Flash | 61.3% | Aug 2026 | Official leaderboard | |
| 12 | | 59.9% | Apr 2026 | Official leaderboard | |
| 13 | GLM-5.1 | 58.5% | Apr 2026 | Official leaderboard | |
| 14 | | 56.4% | Aug 2025 | Official leaderboard | |
| 15 | | 56.3% | Sep 2025 | Official leaderboard | |
| 16 | GLM-4.6 | 54.2% | Sep 2025 | Official leaderboard | |
| 17 | | 53.6% | Mar 2026 | Official leaderboard | |
| 18 | | 52.8% | Apr 2026 | Official leaderboard | |
| 19 | | 51.4% | Mar 2025 | Official leaderboard | |
| 20 | Jamba 1.7 Large | 51.2% | Jul 2025 | Official leaderboard | |
| 21 | | 50.4% | Feb 2026 | Official leaderboard | |
| 22 | | 49.9% | Mar 2025 | Official leaderboard | |
| 23 | Claude Haiku 4.5 | 49.9% | Oct 2025 | Official leaderboard | |
| 24 | gpt-oss-120b | 48.4% | Aug 2025 | Official leaderboard | |
| 25 | | 46.6% | Mar 2026 | Official leaderboard | |
| 26 | Kimi K2 Sep 2025 | 45.7% | Sep 2025 | Official leaderboard | |
| 27 | | 45.3% | Oct 2024 | Official leaderboard | |
| 28 | GPT-4o mini | 43.7% | Jul 2024 | Official leaderboard | |
| 29 | GPT-5 nano | 41.8% | Aug 2025 | Official leaderboard | |
| 30 | | 41% | Dec 2024 | Official leaderboard | |
| 31 | | 31.3% | Dec 2025 | Official leaderboard | |
| 32 | Jamba 1.7 Mini | 30.8% | Jul 2025 | Official leaderboard | |
| 33 | | 29.9% | Jul 2025 | Official leaderboard |
Rankings by part
We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Translation, Summarization and Trivia. Gemini 3 Pro leads 3 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Translation
33 models we track have a score here.
- 1
Gemini 3 Pro 51.2%
- 2
Gemini 3 Flash 50.7%
- 3
Gemini 2.5 Pro Jun 2025 ↑1 50.5%
- 4 GPT-5.1 ↓1 44.2%
- 5
Qwen3.5 397B-A17B 42.9%
- 6 GPT-5 mini ↑1 41.5%
- 7 GLM-5.3-Flash ↑4 40.3%
- 8 GLM-5.1 ↑5 37.4%
- 9
Gemini 2.5 Flash Sep 2025 ↓3 35%
- 10 Claude Sonnet 4.5 ↓2 33.8%
Summarization
33 models we track have a score here.
- 1 GPT-5.1 ↑2 66%
- 2 GLM-5.3-Flash ↑9 60.5%
- 3
MiniMax-M2.7 ↑14 56.5%
- 4 GPT-5 mini ↑3 55.1%
- 5 Claude Sonnet 4.5 ↑3 55.1%
- 6
Nemotron 3 Super ↑19 52.8%
- 7
Gemini 3 Pro ↓6 51.8%
- 8
Gemini 3 Flash ↓6 49.7%
- 9
Gemini 2.5 Pro Jun 2025 ↓5 49.1%
- 10
MiniMax-M2.5 ↑11 48.8%
Trivia
33 models we track have a score here.
- 1
Gemini 3 Pro 86.7%
- 2
Gemini 3 Flash 86.4%
- 3
Gemini 2.5 Pro Jun 2025 ↑1 82.1%
- 4 GPT-5.1 ↓1 81.7%
- 5 GPT-4o Mar 2025 ↑5 77.1%
- 6 GLM-5.1 ↑7 76.7%
- 7 GLM-5.3-Flash ↑4 76.7%
- 8
Gemini 2.5 Flash Sep 2025 ↓2 76.4%
- 9 gpt-oss-120b ↑15 74.8%
- 10
Qwen3.5 397B-A17B ↓5 72.1%
Winograd
33 models we track have a score here.
- 1
Gemini 3 Pro 96%
- 2
Gemini 3 Flash 94.6%
- 3 GLM-5.3-Flash ↑8 93.9%
- 4
Qwen3.5-122B-A10B ↑5 93.2%
- 5 GLM-5.1 ↑8 91.4%
- 6 GPT-5.1 ↓3 91%
- 7
Gemini 2.5 Pro Jun 2025 ↓3 90.6%
- 8 GPT-4o Mar 2025 ↑2 89.2%
- 9 Claude Sonnet 4.5 ↓1 88.8%
- 10 GPT-5 mini ↓3 88.5%
Nikud
33 models we track have a score here.
- 1
Gemini 3 Flash ↑1 88.4%
- 2
Gemini 3 Pro ↓1 87.6%
- 3
Gemini 2.5 Pro Jun 2025 ↑1 87.2%
- 4 GPT-5.1 ↓1 80.5%
- 5
Gemini 2.5 Flash Sep 2025 ↑1 79.5%
- 6
Qwen3.5 397B-A17B ↓1 78.2%
- 7
DeepSeek-V3.1 ↑7 75.3%
- 8 GPT-4o Mar 2025 ↑2 74.8%
- 9
Gemma 4 31B IT ↑3 73.4%
- 10
Gemini 2.5 Flash-Lite Sep 2025 ↑5 70.7%
3 models have held number one on the official leaderboard since its release in November 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Hebrew LLM Chat Leaderboard 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 Hebrew LLM Chat Leaderboard 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
9 of 10 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (34 entries).
- Pass Models scored 33 models we track, plus 1 other entry on its leaderboard.
- Pass Newest score The newest is from August 2026, for GLM-5.3-Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is GLM-5.3-Flash (#53).
- Pass Headroom The best is 74.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on huggingface.co: 33 rows for models we track and 1 other entry 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.