Artificial Analysis lists Amazon's Nova 2.0 Lite
Artificial Analysis has added Amazon's Nova 2.0 Lite to its tracker in four reasoning settings, each with a 1M-token context window and the same $0.52 listed price per million tokens.
Independent benchmarking firm Artificial Analysis has published a release page for Amazon's Nova 2.0 Lite. It tracks the model as four variants: High, Medium and Low reasoning settings, plus a Non-reasoning mode.
According to the listing, all four variants:
- accept text and image input and produce text output;
- support a 1M-token context window, which Artificial Analysis puts at roughly 1,500 A4 pages;
- are priced at $0.52 per 1M tokens;
- are proprietary, with no weights available.
On latency, Artificial Analysis says the Non-reasoning variant has the lowest time to first answer token of the four, at 1.76 seconds.
At that price, Artificial Analysis puts Nova 2.0 Lite in its $0.15–$1 per 1M tokens bracket. The firm compares proprietary models against other proprietary and open-weights models in the same bracket, using a 3:1 blend of input and output prices. The reasoning variants are also compared against non-reasoning models, while the Non-reasoning variant is compared only with other non-reasoning models.
The page lists capability indexes built from evaluations including AA-Omniscience v1.1, AA-Briefcase v1.1, HLE-Diamond, AutomationBench-AA, GDP.pdf, Terminal-Bench 4.0 and FrontierCode-AA 2.0. The material available does not include Nova 2.0 Lite's actual scores on these indexes, its cost per task, or its rank against competing models. It also does not give a release date or any details from Amazon itself.