Jamba Instruct: benchmark results
Provider: AI21 Labs. Access: API.
Aggregate Rating 1416 ± 19, rank #5384 of 10916 rated models, from 93 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM NaturalQuestions (Closed) | 38.36 | F1 (%) | 51.1 |
| HELM MMLU - Sociology | 89.05 | EM | 44.9 |
| HELM MMLU - Us Foreign Policy | 91 | EM | 44.9 |
| HELM Lite - LegalBench - Function of Decision Section | 36.78 | Exact Match (%) | 42.8 |
| HELM Lite - MATH Level 1 - Number Theory | 46.67 | Equivalent (CoT) (%) | 40 |
| HELM Lite - LegalBench - PROA | 87.37 | Exact Match (%) | 38.3 |
| HELM Lite - LegalBench - Corporate Lobbying | 71.22 | Exact Match (%) | 37.2 |
| HELM Lite - GSM8K | 67 | Exact Match (%) | 36 |
| HELM Lite - WMT 2014 (de-en) | 17.73 | BLEU-4 (%) | 35.6 |
| HELM MMLU - Moral Scenarios | 46.48 | EM | 35.3 |
| HELM Lite - WMT 2014 (hi-en) | 9.94 | BLEU-4 (%) | 33.3 |
| HELM MMLU - Machine Learning | 53.57 | EM | 33.3 |
Interactive version: theaggregate.ai/model?slug=jamba-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.