Nova 2.0 Lite: benchmark results
Amazon's Nova 2.0 Lite, the speed- and cost-optimized model in their Nova 2 line announced December 2025, alongside Nova 2.0 Pro and Nova 2.0 Omni. Provider: Amazon. Released 2025-12-02. Access: API.
Unified ELO 1612 ± 11, rank #396 of 1639 rated models, from 456 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Ko-AgentBench - L1 Single Tool Call | 79.55 | Argument Accuracy (%) | 100 |
| LLM Stats (Multi-Challenge) | 76.6 | Score (%) | 96.9 |
| AGC-Bench - grapheval_ai_researcher | 1.11 | Dataset z-score | 96.3 |
| AA AIME 2025 | 94.33 | Accuracy (%) | 95.4 |
| Nejumi 4 - Toxicity - Violation Categories | 57.14 | Criteria met (%) | 95.4 |
| AGC-Bench - hypobench | 0.42 | Dataset z-score | 93.9 |
| Nejumi 4 - Toxicity - Prohibited Acts | 97.92 | Criteria met (%) | 93.6 |
| PRISM-VLM - Sycophancy Resistance | 78.3 | Per-axis mean score (%; share of turn-0-correct answers held | 92.7 |
| Evals for Every Language - Danish (da) | 76.8 | Average Score (%) | 90.9 |
| Nejumi 4 - MT-Bench (Japanese) - Humanities | 100 | Judge rating (1-10, x10) | 90.8 |
| Vectara Hallucination Leaderboard | 94.9 | Factual Consistency Rate (%) | 90.7 |
| Evals for Every Language - Belarusian (be) | 71.82 | Average Score (%) | 90.6 |
Interactive version: theaggregate.ai/model?slug=nova-2-0-lite · How It Works · Data refreshed daily, snapshot 2026-10-09.