AWS

AWS has added four multimodal evaluators to its Strands Evals SDK that judge image-to-text AI outputs by directly analyzing source images. The evaluators—Overall Quality, Correctness, Faithfulness, and Instruction Following—use multimodal large language models to detect visual hallucinations, factual errors, and instruction violations that text-only judges miss.

Amazon SageMaker AI has launched bidirectional streaming support for real-time inference, enabling WebSocket-based voice applications through vLLM integration. The feature uses HTTP/2 on port 8443 to bridge client connections with vLLM's Realtime API, allowing audio to stream in while transcription streams back simultaneously over a single persistent connection.

AWS released Amazon Nova Sonic, a speech-to-speech foundation model for voice agents, alongside AgentCore Runtime, a serverless hosting environment with WebSocket streaming and microVM isolation. The framework supports three integration patterns: direct tool calls via AgentCore Gateway using Model Context Protocol (MCP), sub-agent delegation with Agent-to-Agent (A2A) protocol, and session segmentation for multi-step workflows.

Amazon Bedrock AgentCore Browser now supports Chrome enterprise policies and custom root CA certificates, giving organizations control over 450+ browser settings for AI agents. The feature enables URL filtering, download restrictions, password manager controls, and connectivity to internal services through custom certificate authorities.

AWS has integrated WebRTC protocol support with Amazon Nova Sonic, its speech-to-speech model, through Amazon Kinesis Video Streams. The integration delivers real-time voice streaming with sub-second latency and includes adaptive bitrate control, forward error correction, and Voice Activity Detection for mobile and IoT applications.

AWS announced general availability of Claude Platform on AWS, enabling direct access to Anthropic's native APIs, tools, and console through existing AWS accounts. The service includes the Messages API, Claude Managed Agents, web search, MCP connector, and code execution, authenticated via AWS IAM and billed through AWS Marketplace.

AWS announced Amazon Bedrock AgentCore Payments (preview), enabling AI agents to autonomously discover and pay for APIs, web content, MCP servers, and other agents. Built with Coinbase and Stripe, the service supports micropayments through the x402 protocol with per-session spending limits and full transaction observability.

AWS announced OS Level Actions for Amazon Bedrock AgentCore Browser, extending agent automation beyond the browser's Document Object Model to interact with native operating system UI. The feature enables agents to control system dialogs, security prompts, and context menus through direct mouse and keyboard commands at the OS level.

Amazon Web Services released AgentCore Optimization in preview, introducing an automated performance loop that generates configuration recommendations from production traces, validates them through batch evaluation and A/B testing, and enables continuous agent optimization. The system targets the quality drift problem where AI agents degrade as models evolve and user behavior shifts.

Amazon Web Services has released agent-guided workflows in SageMaker AI that use AI coding agents to automate model customization. The feature includes nine pre-built skills covering use case definition, data preparation, fine-tuning technique selection (SFT, DPO, RLVR), evaluation, and deployment to Amazon Bedrock or SageMaker endpoints.

Amazon SageMaker AI now supports capacity-aware instance pools that automatically try alternative GPU instance types when primary choices lack capacity. The feature works across endpoint creation, autoscaling, and scale-in operations, eliminating the manual retry cycles that previously left endpoints stuck in failed states.

AWS announced that Amazon Q Developer IDE plugins and paid subscriptions will reach end of support on April 30, 2027, with new account creation blocked starting May 15, 2026. The company is transitioning users to Kiro, a new agentic development environment built for spec-driven development.

AWS has released support for custom Model Context Protocol (MCP) proxies on Amazon Bedrock AgentCore Runtime, allowing organizations to implement custom governance and security controls on AI agent tool interactions without modifying upstream MCP servers. The serverless proxy runs on AgentCore Runtime with automatic scaling and built-in observability through CloudWatch and OpenTelemetry.

AWS has added OpenAI's latest models, Codex, and a new managed agents service to its Bedrock platform, one day after OpenAI revised its agreement with Microsoft. The integration follows OpenAI's up-to-$50 billion deal with Amazon.

Amazon Web Services released technical guidance for migrating text agents to voice assistants using Amazon Nova 2 Sonic, a native speech-to-speech model that combines automatic speech recognition, reasoning, tool calling, and text-to-speech in a single bidirectional streaming interface. The model supports asynchronous tool calling and built-in voice activity detection for handling interruptions.

NVIDIA has launched Nemotron 3 Nano Omni on Amazon SageMaker JumpStart, a multimodal model with 30 billion total parameters (3 billion active) that processes video, audio, images, and text in a single inference pass. The model features a 131K token context window and uses a Mamba2 Transformer Hybrid MoE architecture combining three specialized encoders.

Anthropic announced Claude Cowork is now available on Amazon Bedrock, allowing organizations to deploy the desktop AI assistant through their AWS infrastructure with consumption-based pricing. Unlike Claude Enterprise, pricing flows through existing AWS agreements with no per-seat licensing from Anthropic.

Amazon Web Services released a model distillation pipeline on Amazon Bedrock that transfers video search routing intelligence from Nova Premier to Nova Micro. According to AWS, the approach reduces inference cost by over 95% and latency by 50% compared to using Claude Haiku for intent routing.

Amazon released Nova Multimodal Embeddings on Amazon Bedrock, a unified embedding model that processes text, documents, images, video, and audio into a shared 1024-dimensional semantic vector space. The model supports up to 30 seconds of video per embedding and enables semantic search across all modalities simultaneously without converting video to text first.

Amazon Web Services published a practical guide for fine-tuning Amazon Nova models using the Nova Forge SDK's data mixing capabilities. According to AWS, blending customer data with Amazon-curated datasets preserved near-baseline MMLU scores while delivering a 12-point F1 improvement on a Voice of Customer classification task spanning 1,420 leaf categories.

AWS has enabled fine-tuned Amazon Nova Micro models to run on Bedrock's on-demand inference for text-to-SQL generation. According to AWS testing, a sample workload of 22,000 queries per month costs $0.80 monthly using the serverless approach, compared to higher costs with persistent model hosting. The solution uses LoRA fine-tuning on the sql-create-context dataset containing over 78,000 SQL examples.

AWS has released Automated Reasoning checks in Amazon Bedrock Guardrails, a feature that uses formal mathematical verification to validate AI outputs against defined rules. Unlike LLM-as-a-judge approaches that use one probabilistic model to validate another, Automated Reasoning provides mathematically proven, auditable compliance evidence for regulated industries.

AWS has introduced Lambda-based reward functions for Amazon Nova model customization through reinforcement fine-tuning (RFT). The serverless architecture automatically scales from 10 concurrent evaluations per second during experimentation to 400+ during production training, supporting both objective RLVR and subjective RLAIF approaches.

Amazon Bedrock now enables fine-tuning of Amazon Nova models using supervised fine-tuning (SFT), reinforcement fine-tuning (RFT), and model distillation. The service automates infrastructure provisioning and training orchestration, requiring only data upload to S3 and a single API call. Fine-tuned models run on-demand at standard inference pricing without provisioned capacity requirements.

Amazon Bedrock AgentCore Runtime now offers managed session storage to persist agent filesystem state across stop/resume cycles and InvokeAgentRuntimeCommand for executing shell commands directly within agent microVMs. The features address two core challenges in production agent workflows: ephemeral filesystems that reset between sessions and the inability to execute deterministic operations without routing them through LLMs.