[whatmodel.sh](../../README.md) / Nova 2 Pro (Preview) # Nova 2 Pro (Preview) Amazon · released 2025-11 · closed weights **score 62.9** ±1.9 · #173 of 732 · from 3 benchmark results across 2 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at _{● marks Nova 2 Pro (Preview) among every model on that task or test, worst to best.}

task         vs the others          rank  counts  tests
agent tasks  ▁▁▃▆█▄▅●▄▄▂▁   #101 of 313*    7.4%      1

agent tasks · #101 of 313 · 7.4% of its score

Beats [Grok 4.20 (0309 reasoning)](grok-4-20.md), [o3-pro](o3-pro.md), [Grok 4.3](grok-4-3.md) here, though they rank higher overall.

test                                result  vs the others in its run  difficulty        counts  why
τ²-Bench-Verified (GPT-5.1 user) …   78.5%  ▂··▄···▂▄▄●▄ beats 8/13   medium · avg 45%    7.4%  at its level

other tests (2)

test                                result  vs the others in its run  difficulty           counts  why
KMMLU-HARD (Daekeun Kim, zero-sho…   54.1%  █···▁▁▁▂●·▆▂ beats 14/23  very hard · avg 29%     46%  sharp · at its level
CLIcK Korean (Daekeun Kim)           73.1%  ▆▄▄▄●▄▂·▄█·▄ beats 12/23  medium · avg 56%        45%  typical

## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md)

source                                   results  share of the score
Daekeun Kim Korean LLM benchmark runs          2   91% ▇▇▇▇▇▇▇▇▇▇▇
Amazon AGI (submitted / maker-reported)        1    7% ▇
prior: pull toward the middle                       2% ▇
weighted average                                  = 62.9

_{Licences: Daekeun Kim Korean LLM benchmark runs ([MIT](../../licenses/korean-personal-benchmark-runs-MIT.txt)) · Amazon AGI ([MIT](../../licenses/tau2-verified-MIT.txt))} ## How to run it ### Through an API api (Amazon Bedrock) · $3.78/M blended · $1.375 in / $11 out per 1M tokens · [aws.amazon.com](https://aws.amazon.com/nova/pricing/) _{[price source](https://aws.amazon.com/nova/pricing): USD per 1M text tokens, Amazon Bedrock Standard tier, geo cross-Region inference and in-Region, Amazon Nova 2 Pro (Preview). Global cross-Region inference is listed lower. Page table values are rendered from AWS's own price data file for the page (b0.p.awsstatic.com meteredUnitMaps, bedrock USD, published 2026-09-25T06:55:02Z); US East (N. Virginia) selected. Priority/Flex/Batch tiers, cached tokens and non-text modalities excluded. Currency: USD. AWS official Amazon Nova pricing page, observed 2026-09-25. · [context source](https://github.com/BerriAI/litellm/blob/c19ce71bcdad234ca87a7af10cee45375ce811a8/model_prices_and_context_window.json): LiteLLM (MIT, pinned commit c19ce71) Amazon Bedrock row amazon.nova-2-pro-preview-20251202-v1:0 (also its us./eu./apac. cross-Region ids): max\_input\_tokens 1,000,000, max\_output\_tokens 64,000. Provider maximum input tokens on Bedrock, the route this model is priced on. models.dev @a668908 (providers/nova/models/nova-2-pro-v1.toml) also gives context 1,000,000 for Nova 2 Pro. AWS's Nova 2 user guide says 'All models support up to 1 million tokens of context' but its model table lists only Nova 2 Lite, Sonic and Multimodal Embeddings. Observed 2026-09-26.} [interactive version](https://whatmodel.sh/model/nova-2-pro) · [JSON](https://whatmodel.sh/api?format=json&all=1)