source
Scale Labs
Benchmarks · /benchmarks/scale-multinrc
MultiNRC
Scale multilingual reasoning benchmark across languages and reasoning-heavy prompts.
Passport
Visible tradeoffsThis is a rubric-judged signal, so it is more structured than arena taste but still depends on the scoring rubric.
metric
Score (%)
judge
Rubric
direction
higher better
group id
scale_multinrc_current
domain
Reasoning / math / science
What it measures vs what it misses
✓ Measures
Cross-lingual reasoning and answer quality on multilingual tasks.
✗ Misses
Search freshness. Image-grounded reasoning.
Why this countsIt is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.Comparable-group ruleThis percentile only compares models inside the exact benchmark/version group shown here. It is not a universal score.What it missesIt still misses product usability, latency, and whether the model stays correct in messy real workflows.
Leaderboard · this benchmark version
#1 · Gemini 3.1 Pro
SL · Apr 29, 2026
64.7%
#2 · GPT-5.4
SL · Apr 29, 2026
62.3%
#3 · Gemini 3 Pro Preview
SL · Apr 29, 2026
59%
#4 · Claude Opus 4.6
SL · Apr 29, 2026
57.1%
#5 · GPT-5
SL · Apr 29, 2026
52.1%
#6 · GPT-5.4 mini
SL · Apr 29, 2026
52.1%
#7 · GPT-5.4 nano
SL · Apr 29, 2026
52.1%
#8 · GPT-5.1
SL · Apr 29, 2026
49%
#9 · o3
SL · Apr 29, 2026
45.5%
#10 · Gemini 2.5 Pro
SL · Apr 29, 2026
45.1%
#11 · DeepSeek Reasoner
SL · Apr 29, 2026
27.6%