Nova Pro

Nova Pro sits in the middle of Amazon's Nova model family, designed to handle multimodal inputs including text, images, and video while producing text outputs. The model targets developers and businesses that need robust tool-calling capabilities and precise temperature control for customizing response characteristics. Its architecture reflects Amazon's investment in building a capable generalist model that can serve as the backbone for a wide range of production applications without requiring the largest or most expensive frontier model.

Benchmark results published through Amazon Science reveal the model's particular strengths: Nova Pro achieved a 0.95/1 score on the ARC-Challenge commonsense reasoning benchmark and matched that score on the Grade School Math 8K benchmark, both evaluated using 0-shot Chain-of-Thought methodology. The model demonstrates strong performance in tool calling (ranked 83 on LLM Leaderboard) and solid capabilities across finance (85), legal (99), and vision tasks (116). With a 300,000-token context window, it can process lengthy documents or conversation histories, making it well-suited for enterprise workflows involving complex documents, multi-step agentic tasks, and applications that require the model to maintain coherence across long interactions. The December 2024 release timing positions it as a mainstream production option for teams seeking capable multimodal reasoning without frontier-model costs.