AWS Bedrock adds Claude Opus 5.5, Grok 4.6, and Moonshot's 2.8T-parameter Kimi K3

AWS used the week's launch wave to broaden Bedrock into the definitive multi-model marketplace, folding in same-day availability of Claude Opus 5.5, OpenAI's GPT-6 Sol and Luna, and xAI's Grok 4.6. The pitch to enterprises is unified controls: one set of Bedrock APIs, security, governance, and cost management across models from competing labs.

The most strategically notable addition is Moonshot AI's Kimi K3, an open-source model with 2.8 trillion parameters, which launches under a revenue-sharing partnership. That deal is a template for how Chinese labs can find overseas distribution despite geopolitical friction — Moonshot gets a compliant Western channel, AWS gets a competitive open-weight option, and both share revenue. Developers on the community side read it as evidence that open-weight momentum is real and increasingly cross-border.

Mechanically, the value is that teams can match intelligence, speed, and cost to each workload without re-plumbing infrastructure: GPT-6 Luna for cheap high-volume tasks, Opus 5.5 for agentic coding, Grok for coding autonomy, Kimi K3 for open-weight flexibility. AWS also tied these into its agent tooling (AgentCore, Strands) shipped the same week.

The competitive context is Bedrock versus Azure Foundry, which also added GPT-6 Astra, Sol, and Luna this week. AWS's differentiator is breadth — it is the only major cloud carrying Anthropic, OpenAI, xAI, and Chinese open-weight models simultaneously. The caveat for buyers is that 'available on Bedrock' spans wildly different price/performance tiers, and the same-day rush means real-world benchmarking on these newest models is still thin.