Benchmark Almanac

Nova 2 Lite

Nova 2 Lite ranks first on QVHighlights.

Amazon Released December 2nd, 2025 Family Amazon Nova 2 The lab's model card Compare with others
28
Benchmarks with a score
1
Ranked first, out of five or more models
2
In the top three
Amazon Dec 2025

Nova 2 Lite

Closed weights Mid-size

139.1 #182 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • MMMU-Pro 61.8% #65 /90
  • IFBench 70.8% #43 /80
  • BFCL V4 60.3% #20 /30

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
MultiChallenge 76.6% #2 of 42 Lab model card
Vectara Hallucination Leaderboard (2025 dataset) 5.10% #12 of 104 Official leaderboard
MMLU-Pro 80.9% #72 of 167 Lab model card
IFBench 70.8% #43 of 80 Lab model card
LMArena (Elo) 1,335 #196 of 335 Official leaderboard
StudyArena Text 1,191 #67 of 85 Official leaderboard
Longform Creative Writing (EQ-Bench) 38.8 #101 of 119 Official leaderboard
LongCodeBench 84% #2 of 2 Lab model card

Math and science

BenchmarkScoreRankSource
AIME 2025 91% #45 of 143 Lab model card
GPQA Diamond 79.6% #121 of 302 Lab model card

Coding

BenchmarkScoreRankSource
LiveCodeBench v5 71% #8 of 47 Lab model card
SWE-bench Verified 64.5% #100 of 155 Lab model card

Agents and tools

BenchmarkScoreRankSource
tau2-bench 76% #48 of 79 Lab model card
BFCL V4 60.3% #20 of 30 Lab model card
PinchBench 63.1% #37 of 48 Official leaderboard
Terminal-Bench 32.5% #18 of 22 Lab model card
tau2-bench Verified 72.4% #9 of 11 Official leaderboard
ScreenSpot 83.3% #10 of 11 Lab model card
PinchBench 2.0 37.6% #53 of 57 Official leaderboard
MCP Atlas 24.6% #46 of 46 Lab model card

Vision, audio and robotics

BenchmarkScoreRankSource
QVHighlights 77.2% #1 of 5 Lab model card
RealKIE-FCC Verified 62.1% #2 of 3 Lab model card
MMMU-Pro 61.8% #65 of 90 Lab model card
ParseBench 40.4% #38 of 44 Official leaderboard
OCRBench v2 56.1% #19 of 22 Lab model card

Safety and security

BenchmarkScoreRankSource
Aegis 2.0 85.8% #1 of 3 Lab model card
Jigsaw Toxic Comment Classification 56.5% #2 of 3 Lab model card
WildGuardTest 84.7% #4 of 5 Lab model card

Its lab's numbers put it at 139.1, number 182 of 404

The Self-Reported Index fits one score to the 14 capability results Amazon reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 135.9 to 142.4. How we compute it.

Its lab's numbers next to 2 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
BFCL V4 60.3% BFCL v4 Overall FC 27.1% Leaderboard, native function calling +33 pts
tau2-bench Verified 70.7% Mean of τ2-bench Retail-Verified 76.5 / Airline-Verified 64.8 (avg@3) 72.4% Leaderboard, tau2-bench-verified; gpt-5.1 user simulator −1.7 pts

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Amazon Nova 2: Multimodal Reasoning and Generation Models (Technical Report and Model Card) Dec 2, 2025

Its system card

Every result in the longer documents the lab published with it. 4 of them aren't in the launch post.