Frontier AI
Max 5x gets $100, Max 20x $200, and Team $20 or $100 per seat, pooled up to $500, after linking a Console organization.
Why it matters. Credits cover the API, Managed Agents and the Agent SDK but not Claude Code, and unused credits expire each billing cycle.
Version 2.1.295 adds onFailure: "block", so a command or HTTP hook that cannot start, times out or exits unexpectedly stops the action.
Why it matters. Set it on guard hooks so a crashed or slow check blocks the tool call instead of letting it through.
Source: Claude Code v2.1.295 release
With the worktrees feature enabled, Codex gets tools to create and list managed Git worktrees from trusted local projects.
Why it matters. Parallel agent tasks can each work in their own checkout instead of editing the same working tree.
Source: Codex CLI 0.162.0 release
In beta, you subclass a toolset class and write one method per action, such as navigate or left_click, against your own automation.
Why it matters. The SDK runs the tool loop, your URL and file policies and your approval callback, but you still supply the browser or desktop.
Mellum2.1 is a 12B mixture of experts model with 2.5B active parameters, released under Apache 2.0 on Hugging Face.
Why it matters. It targets fast subagents on your own hardware, and GGUF builds for llama.cpp, Ollama and LM Studio are listed as coming soon.
Source: JetBrains AI blog