Benchmark Almanac

MAXIFE

Qwen3.7-Max 89.2% leads since May 2026
Fading

Models follow 1,667 verifiable instructions in 23 languages and in cross-lingual settings. Rule-based checks and model judges score compliance.

Qwen model cards report accuracy on English plus multilingual original prompts, 23 settings in total.

Released
June 2025
Kind
Chat and instructions in many languages
Level
Everyday
Scored
Checked automatically
Answers
Open-ended
Tasks
All public
Strongest model tested
Qwen3.7-Max #49 on the capability index
Last in a launch post
June 2026, dropped after it saturated
Last model added
June 2026

In the authors’ words

With the rapid adoption of large language models (LLMs) in natural language processing, the ability to follow instructions has emerged as a key metric for evaluating their practical utility. However, existing evaluation methods often focus on single-language scenarios, overlooking the challenges and differences present in multilingual and cross-lingual contexts. To address this gap, we introduce MaXIFE: a comprehensive evaluation benchmark designed to assess instruction-following capabilities across 23 different languages with 1667 verifiable instruction tasks. MaXIFE integrates both Rule-Based Evaluation and Model-Based Evaluation, ensuring a balance of efficiency and accuracy. We applied MaXIFE to evaluate several leading commercial LLMs, establishing baseline results for future comparisons. By providing a standardized tool for multilingual instruction-following evaluation, MaXIFE aims to advance research and development in natural language processing.

Abstract of MaXIFE: Multilingual and Cross-lingual Instruction Following Evaluation (arXiv:2506.01776, CC0)
11
Models scored in lab cards
11
Launch posts list it, from 1 lab

Qwen3.7-Max leads with 89.2% among models released by June 2026

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

The newest model with a score here came out in June 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 89.2 Qwen3.7-Max
  2. 88.8 Qwen3.7-Plus
  3. 88.2 Qwen3.5 397B-A17B
  4. 88.2 Qwen 3.6 Plus
  5. 88.0 Qwen3.5-27B
  6. 87.9 Qwen3.5-122B-A10B
  7. 86.6 Qwen3.5-35B-A3B
  8. 83.4 Qwen3.5-9B
  9. 78.0 Qwen3.5-4B
  10. 60.6 Qwen3.5-2B
  11. 39.2 Qwen3.5-0.8B
  1. 1 Qwen3.7-Max 89.2%
  2. 2 Qwen3.7-Plus 88.8%
  3. 3 Qwen3.5 397B-A17B 88.2%
  4. 4 Qwen 3.6 Plus 88.2%
  5. 5 Qwen3.5-27B 88%
  6. 6 Qwen3.5-122B-A10B 87.9%
  7. 7 Qwen3.5-35B-A3B 86.6%
  8. 8 Qwen3.5-9B 83.4%
  9. 9 Qwen3.5-4B 78%
  10. 10 Qwen3.5-2B 60.6%
  11. 11 Qwen3.5-0.8B 39.2%

Source: Labs’ model cards

#ModelScoreSettingDateSource
1 Qwen3.7-Max 89.2% May 2026 Lab model card
2 Qwen3.7-Plus 88.8% Jun 2026 Lab model card
3 Qwen3.5 397B-A17B 88.2% Feb 2026 Lab model card
4 Qwen 3.6 Plus 88.2% Apr 2026 Lab model card
5 Qwen3.5-27B 88% Feb 2026 Lab model card
6 Qwen3.5-122B-A10B 87.9% Feb 2026 Lab model card
7 Qwen3.5-35B-A3B 86.6% Feb 2026 Lab model card
8 Qwen3.5-9B 83.4% Mar 2026 Lab model card
9 Qwen3.5-4B 78% Mar 2026 Lab model card
10 Qwen3.5-2B 60.6% Mar 2026 Lab model card
11 Qwen3.5-0.8B 39.2% Mar 2026 Lab model card

2 models have held number one in labs’ own numbers since its release in June 2025

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

Qwen
Lab claims Passed 85% in Feb 2026, by Qwen3.5 397B-A17B
Qwen3.7-Max 89.2%

Source: Labs’ model cards

Qwen’s own number for Qwen3.5 397B-A17B passed 85% in February 2026

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

Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Qwen3.7-Max Qwen3.5 397B-A17B 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Qwen3.7-Max

Source: Labs’ model cards

Labs headline it less often now, last in June 2026

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Qwen
DateLaunch postScoreAs the post writes it
Jun 1, 2026 Qwen3.7-Plus 88.8 MAXIFE
May 20, 2026 Qwen3.7-Max 89.2 MAXIFE
Apr 2, 2026 Qwen 3.6 Plus 88.2 MAXIFE
Mar 2, 2026 Qwen3.5-0.8B 39.2 MAXIFE
Mar 2, 2026 Qwen3.5-2B 60.6 MAXIFE
Mar 2, 2026 Qwen3.5-4B 78 MAXIFE
Mar 2, 2026 Qwen3.5-9B 83.4 MAXIFE
Feb 24, 2026 Qwen3.5-35B-A3B 86.6 MAXIFE
Feb 24, 2026 Qwen3.5-27B 88 MAXIFE
Feb 24, 2026 Qwen3.5-122B-A10B 87.9 MAXIFE
Feb 16, 2026 Qwen3.5 397B-A17B 88.2 MAXIFE

More in Multilingual

All 120 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. Nemotron 3 Ultra 550B-A55B Base 90.1%
  • 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%