Amazon launches second generation of its Nova AI models, including Nova Lite, Nova Pro, Nova Sonic, and Nova Omni, a fully multimodal reasoning model
Will Knight / Wired :
Context & Ripple Effects
Nova’s second generation consolidates work Amazon had already signaled: a planned hybrid-reasoning Nova model and the earlier release of Nova Sonic for natural-sounding speech. Adding Omni extends that lineup from discrete model types toward a single multimodal reasoning offering.
The release also fits Amazon’s stated Alexa+ approach of selecting among Nova and Anthropic models by task. A broader in-house Nova portfolio gives that model-selection strategy more internal options, even as it remains model agnostic.
First-order effects
- Amazon’s Nova lineup now spans Lite, Pro, Sonic, and Omni, with Omni positioned as a fully multimodal reasoning model; developers and Amazon product teams gain a more segmented set of first-party model choices.
- The launch turns the earlier reasoning-model plan into a product-family expansion, while placing the existing voice-focused Sonic model inside a broader Nova generation.
Second-order effects
- Amazon can evaluate a wider range of tasks against its own models before relying on external providers, reinforcing the operational value of its model-agnostic routing approach for Alexa+.
- Competing model providers will face a more complete Amazon-owned alternative across lightweight, higher-end, voice, and multimodal reasoning workloads, making capability and deployment trade-offs more central to model selection.
Third-order effects
- If Amazon continues to broaden Nova by modality and capability tier, cloud and consumer-platform operators may increasingly compete through integrated model portfolios rather than a single flagship model.
- The larger shift is toward AI stacks in which firms combine proprietary and third-party models, routing work by task while using owned models to improve cost control and product differentiation.
The trend: This is one data point in the rise of vertically integrated AI platforms that build model families for different workloads while retaining the flexibility to mix in outside models.