AWS adds OpenAI models after Microsoft exclusivity ends

AWS announced OpenAI model offerings including a new agent service one day after Microsoft agreed to end exclusive rights. The launch includes GPT-4o and other OpenAI models available through AWS infrastructure.

Platform engineers should evaluate AWS pricing against Azure OpenAI Service for existing workloads. Test the new agent service APIs this week to compare latency and cost per call.

NVIDIA released Nemotron 3 Nano Omni, a single model that handles screens, documents, audio, video, and text for agent workflows. The model is designed for perception-to-action loops without requiring multiple specialized models.

AI engineers building multimodal agents should request early access to test against current multi-model pipelines. Benchmark inference costs for your specific media mix before committing to architecture changes.

Google signed a new contract expanding Pentagon access to its AI models after Anthropic refused to allow DoD use for domestic mass surveillance and autonomous weapons. The contract covers Google's Gemini models for defense applications.

Engineering leaders at defense contractors should audit current model usage policies against new Google terms. Legal teams should review vendor agreements for acceptable use policy changes that could affect compliance.