Reader-focused coverage of AI models, APIs, infrastructure, developer tools and changes that matter in practice. Client service announcements remain inside the authenticated account area.
Z.ai’s GLM-5.3 is now listed on OpenRouter with OpenAI-compatible access, a 1M-token context window and published token pricing. For gateways and coding-agent teams, the release adds another long-context reasoning model to test — but direct Z.ai API availability still needs separate confirmation.
Read news →DeepSeek’s V4 API pricing now includes peak and off-peak billing, adding a time dimension to model-cost planning. For gateways, agents and high-volume teams, the change makes scheduling and cache economics part of routing strategy.
Read news →OpenAI is previewing GPT-5.6 Sol Ultrafast in the API, promising much lower latency through Cerebras-powered inference. Access is limited, but the move gives AI API gateways and enterprise teams a new routing dimension to manage: speed tier eligibility.
Read news →Google’s Imagen 4 standard, ultra and fast Gemini API endpoints are scheduled to shut down on August 17, 2026. Apps and gateways still pinned to those model IDs need routing changes before image requests begin failing.
Read news →GitHub has made Agent Plugins 1.0 generally available across key Copilot surfaces, bringing a shared packaging format for agent skills and MCP servers into VS Code, Copilot CLI, the Copilot SDK and the Copilot app.
Read news →DeepSeek V4-Pro is now listed for API use with a 1M-token context window, very large output limits, OpenAI- and Anthropic-format endpoints, and published token prices that make cache behavior central to cost planning.
Read news →Google has introduced Gemini 3.7 Flash for developers and enterprises, with Gemini API availability and introductory token pricing through the end of 2026. For teams running coding agents and high-volume workflows, the launch is as much a pricing event as a model event.
Read news →SpaceXAI has released Grok 4.6 with API access, OpenAI-compatible chat examples, a 500K context window and published $2/$6 per-million-token pricing, giving developers and AI gateways another frontier model to evaluate for long-context agent and coding workloads.
Read news →Notion Workers now require Notion credits after a free beta period, turning background workspace automations into a metered resource that admins need to monitor alongside Custom Agents and Autofill.
Read news →OpenAI has split Daybreak access into Blue and Red tiers and added GPT-5.6-Cyber for approved defenders, creating a more formal access model for cybersecurity AI workflows in the API.
Read news →GitHub has made Kimi K3 generally available in Copilot with usage-based billing, adding another model choice to developer workflows and reinforcing the need for enterprise policy, cost visibility and model-selection controls.
Read news →AWS Agent Registry is moving out of the Bedrock AgentCore public-preview namespace and into its own agent-registry namespace, requiring updates to endpoints, IAM policies, SDK clients, CLI scripts and registry data.
Read news →