Nova Pro: benchmark results

Amazon's mid-tier Nova multimodal model handling text, image and video input with a 300K context on Bedrock (December 2024). Provider: Amazon. Released 2024-12-03. Access: API.

Unified ELO 1529 ± 1, rank #549 of 1392 rated models, from 102 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELM NaturalQuestions (Open)82.9F1 (%)100
Translation en to Set1 spBleu43.4en→Set1 spBLEU (self-reported)100
HELM (Stanford)88.48Mean Win Rate (%)95.6
HELM NarrativeQA79.13F1 (%)95
Translation Set1 to en spBleu44.4Set1→en spBLEU (self-reported)91.7
HELM WMT 201422.89BLEU-4 (%)90
VIBench12.2Direct VIB (self-reported)88.9
LLM Stats (ChartQA)89.2Score (%)88
ProLLM - Entity Extraction84.5Score (%)87.2
LLM Stats (DROP)85.4Score (%)85.7
Ukrainian Legal Text Benchmark (EDRSR subset)82.73-task Composite (self-reported)83.3
Translation Set1 to en COMET2289Set1→en COMET22 (self-reported)79.2

Interactive version: theaggregate.ai/model?slug=nova-pro · How It Works · Data refreshed daily, snapshot 2026-09-05.