Benchmark Almanac

English to Icelandic Test Suite for Syntactic Divergence

GPT-5.5 93.4% leads since June 2026
Not in model cards

Models translate 345 English sentences into Icelandic. Human evaluators judge whether each translation uses an acceptable Icelandic construction for the targeted English syntax.

The researchers report human scores for 27 systems across 20 constructions. Seven systems have human judgments for all 345 sentences.

Released
June 2026
Kind
Translation
Level
Everyday
Scored
Graded by people
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
English, Icelandic
In and out
Text → Text
Interaction
Single turn
Items from
Written by experts
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
May 2026
Scores refreshed
November 20th, 2026
27
Entries in its paper, 13 models we track

GPT-5.5 leads with 93.4% among models released by May 2026 GPT-5.5 leads with 93.4% among models released by May 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. 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 May 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 GPT-5.5 93.4%
  2. 2 TRIVE 87.9%
  3. 3 Gemini 3.1 Pro 87.2%
  4. 4 Wayfinder 82.7%
  5. 5 SCIR-TG-MT 82.3%
  6. 6 UvA-MT 80.6%
  7. 7 Lumen 80.5%
  8. 8 Tower 9B 79.4%
  9. 9 GoogleTranslate 76.4%
  10. 10 SRPOL 68.6%
  11. 11 DeepSeek-V4-Pro 67.7%
  12. 12 AMI 66.9%
  13. 13 Dubformer 65.4%
  14. 14 SalamandraTA 64.9%
  15. 15 Qingqiu-MT-9B 62.1%
  16. 16 Cohere CAT+ 62%
  17. 17 Gemma 4 31B IT 54.5%
  18. 18 Command A+ 50%
  19. 19 Qwen3.6 27B 47.9%
  20. 20 gpt-oss-120b 45.3%
  21. 21 Mistral Medium 3.5 38.9%
  22. 22 Ministral 3 14B 28.4%
  23. 23 gpt-oss-20b 28.3%
  24. 24 Qwen3.5-9B 27.4%
  25. 25 GLM-4-9B 20.2%
  26. 26 Gemma 4 - 4B 18.4%
  27. 27 Tiny Aya Global 14.2%

Source: The English to Icelandic Test Suite for Syntactic Divergence paper

#EntryScore ±SettingDateSource
1 GPT-5.5 93.4% Apr 2026 Official leaderboard
2 TRIVE 87.9% Official leaderboard
3 Gemini 3.1 Pro 87.2% Feb 2026 Official leaderboard
4 Wayfinder 82.7% ±7 Official leaderboard
5 SCIR-TG-MT 82.3% ±6.6 Official leaderboard
6 UvA-MT 80.6% ±7.1 Official leaderboard
7 Lumen 80.5% ±7.4 Official leaderboard
8 Tower 9B 79.4% ±7.1 Official leaderboard
9 GoogleTranslate 76.4% Official leaderboard
10 SRPOL 68.6% ±8.1 Official leaderboard
11 DeepSeek-V4-Pro 67.7% ±8.9 Apr 2026 Official leaderboard
12 AMI 66.9% Official leaderboard
13 Dubformer 65.4% ±8.4 Official leaderboard
14 SalamandraTA 64.9% ±8.6 Official leaderboard
15 Qingqiu-MT-9B 62.1% ±8.2 Official leaderboard
16 Cohere CAT+ 62% ±8.4 Official leaderboard
17 Gemma 4 31B IT 54.5% Apr 2026 Official leaderboard
18 Command A+ 50% ±9.1 May 2026 Official leaderboard
19 Qwen3.6 27B 47.9% ±8.5 Apr 2026 Official leaderboard
20 gpt-oss-120b 45.3% ±8.5 Aug 2025 Official leaderboard
21 Mistral Medium 3.5 38.9% ±8.8 Apr 2026 Official leaderboard
22 Ministral 3 14B 28.4% ±7.2 Dec 2025 Official leaderboard
23 gpt-oss-20b 28.3% ±7.8 Aug 2025 Official leaderboard
24 Qwen3.5-9B 27.4% ±7.4 Feb 2026 Official leaderboard
25 GLM-4-9B 20.2% Jun 2024 Official leaderboard
26 Gemma 4 - 4B 18.4% ±5.5 Official leaderboard
27 Tiny Aya Global 14.2% ±5 Feb 2026 Official leaderboard
#ModelScore ±SettingDateSource
1 GPT-5.5 93.4% Apr 2026 Official leaderboard
2 Gemini 3.1 Pro 87.2% Feb 2026 Official leaderboard
3 DeepSeek-V4-Pro 67.7% ±8.9 Apr 2026 Official leaderboard
4 Gemma 4 31B IT 54.5% Apr 2026 Official leaderboard
5 Command A+ 50% ±9.1 May 2026 Official leaderboard
6 Qwen3.6 27B 47.9% ±8.5 Apr 2026 Official leaderboard
7 gpt-oss-120b 45.3% ±8.5 Aug 2025 Official leaderboard
8 Mistral Medium 3.5 38.9% ±8.8 Apr 2026 Official leaderboard
9 Ministral 3 14B 28.4% ±7.2 Dec 2025 Official leaderboard
10 gpt-oss-20b 28.3% ±7.8 Aug 2025 Official leaderboard
11 Qwen3.5-9B 27.4% ±7.4 Feb 2026 Official leaderboard
12 GLM-4-9B 20.2% Jun 2024 Official leaderboard
13 Tiny Aya Global 14.2% ±5 Feb 2026 Official leaderboard

Number one over time

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

OpenAI
Benchmark paper Passed 85% in Jun 2026, by GPT-5.5
GPT-5.5 93.4%

Source: The English to Icelandic Test Suite for Syntactic Divergence paper

Gemini 3.1 Pro on the official leaderboard passed 85% in February 2026

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
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) GPT-5.5 Gemini 3.1 Pro gpt-oss-120b GLM-4-9B 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) GPT-5.5 gpt-oss-120b

Source: Official English to Icelandic Test Suite for Syntactic Divergence 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

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 benchmark’s paper (27 entries).
  • Pass Models scored 13 models we track, plus 14 other entries on its leaderboard.
  • Note Newest score The newest came 6 months ago, in May 2026, for Command A+. 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 GPT-5.5 (#21).
  • Note Headroom Saturated. The best is 93.4%, past 85% since February 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on aclanthology.org: 13 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