GovCloud models and developer platform updates #66

Today's Letter

1. AWS Bedrock adds GPT OSS and Nemotron in GovCloud
2. Vercel adds Service Bindings for internal service calls
3. GitHub Copilot adds Kimi K2.7 Code as first open-weight model

AWS Bedrock adds GPT OSS and Nemotron in GovCloud

• Amazon Bedrock in AWS GovCloud (US) now supports OpenAI gpt-oss-120b, gpt-oss-20b, and NVIDIA Nemotron models.
• Added Nemotron variants include Nano 9B v2, Nano 12B v2, Nano 30B, and Super 120B.
• The release targets U.S. government and regulated workloads that need open-weight models inside a controlled boundary.
• Inference runs within the AWS GovCloud (US) isolation boundary on AWS-operated infrastructure in the United States.
• AWS positions the setup for data residency, security, and compliance-sensitive mission systems.
• Cited compliance targets include FedRAMP High, DoD SRG Impact Levels 2, 4, and 5, plus ITAR and CJIS.
• Bedrock exposes the models through one API, allowing model changes without rewriting application code.
• AWS says Nemotron 3 Super uses a 120B MoE design with 12B active parameters per token and a 1M-token context window.

Source: aws.amazon.com

Vercel adds Service Bindings for internal service calls

• Vercel introduced Service Bindings, a feature for secure service-to-service communication within the same deployment.
• A service can declare another service as a binding, and Vercel injects the configured environment variable with the internal URL automatically.
• Application code can call that URL with a normal HTTPS fetch, while Vercel handles internal rewrite, routing, authentication, and TLS setup.
• The feature is aimed at multi-service deployments such as a Next.js frontend calling a FastAPI backend without exposing the backend on a public route.
• Vercel said the traffic stays on its internal network and does not go through the public route table.
• A target service is reachable only when it is exposed, either publicly through a rewrite route or privately through a binding.
• Service-to-service calls appear in Vercel observability, including which bound service was called and the request latency.
• These requests are billed as Service Requests and Fast Origin Transfer, not as CDN Requests or Fast Data Transfer.

Source: vercel.com

GitHub Copilot adds Kimi K2.7 Code as first open-weight model

• GitHub made Kimi K2.7 Code generally available in GitHub Copilot on July 1, 2026.
• GitHub describes it as the first open-weight model selectable in the Copilot model picker, positioned as a lower-cost option for coding workflows.
• The model is hosted by GitHub on Microsoft Azure and billed at provider list pricing under usage-based billing.
• Rollout starts with Copilot Pro, Pro+, and Max, with gradual availability as GitHub monitors quality and performance.
• GitHub said support is available or planned across VS Code 1.127.0+, Visual Studio 17.14.6+, Copilot CLI, the Copilot cloud agent, Copilot App, github.com, GitHub Mobile, JetBrains 1.9.1-251+, Xcode, and Eclipse.
• Copilot Business and Copilot Enterprise keep Kimi K2.7 Code disabled by default until plan administrators enable the model policy in Copilot settings.
• GitHub recommends that organizations review open-weight models against internal security, compliance, and data-governance requirements before enabling access.

Source: github.blog

Jocoletter curates AI, software, and product trends for developers and builders.

#AWS #GitHub #Vercel

Subscribe to Jocoletter