Opus 4.8 Becomes Default, Plus Anthropic Hits $965B

I'm Shannon, and this is the Claude Notes Brief -- Claude Code updates and Anthropic news for the week of June first. Claude Opus four point eight is now the default model. Dynamic workflows can coordinate hundreds of background agents. And Anthropic's valuation now sits at nine hundred sixty five billion dollars.

Let's start with Claude Code, where the headline is a new default model -- Claude Opus four point eight is now what you get out of the box, with high effort turned on automatically. If you've got something even harder, there's a new extra-high effort tier you can reach for. Fast mode on the new Opus is also a notable shift -- it runs at twice the standard rate but delivers about two and a half times the speed, which is a meaningful drop from where fast-mode Opus pricing used to sit. Auto mode, which routes between models for you, also expanded this week to Bedrock, Vertex, and Foundry for Opus four point seven and four point eight.

Building on that model-and-routing theme, dynamic workflows are the other headline feature. You can now ask Claude to create a workflow, and it will plan the work, dispatch tens or even hundreds of background agents in parallel, and verify the results before handing back to you. There's a new command to view your active runs, and the inline progress display got simplified to a single status row under your prompt so the screen doesn't fill up while agents are running. The plugin system also got more flexible -- plugins dropped into a local skills directory are now auto-loaded, no marketplace required, and there's a scaffolding command if you want to spin up a new one from scratch.

And on the resilience side, when your primary model isn't available, Claude Code now automatically falls back to your configured fallback model for the rest of the session instead of failing every request. One small behavior change worth flagging if you script against it -- the simplify command now runs a cleanup-only review focused on reuse and efficiency, rather than a full bug hunt.

Moving under the hood, the theme this week is security fixes and performance on long sessions. On the security side, the team closed a regression where custom API gateways could end up receiving your Anthropic OAuth credential instead of the gateway's own token. Related fixes tightened how subagents handle MCP server configuration, so strict MCP config flags and managed allow and deny policies are now properly honored. And background-session subagents no longer bypass worktree isolation to write into your shared checkout.

On the performance side, long and resumed conversations now render and reload faster, thanks to eliminating redundant work during message rendering. Terminal styling also holds up better in very long sessions, and resuming a session by transcript path no longer eats multiple gigabytes of memory on machines that have a lot of stored sessions. If you live in long-running Claude Code sessions, this is the quiet upgrade of the week.

On to Anthropic news, where the funding story is hard to ignore. The Wall Street Journal is reporting that Anthropic closed a sixty five billion dollar Series H, putting its post-money valuation at nine hundred sixty five billion dollars -- which passes OpenAI and makes Anthropic the most valuable AI startup. We'll link the piece in the show notes. Connected to that capital story, Bloomberg is reporting that Apollo is shopping a thirty six billion dollar debt package to finance Google TPUs dedicated to Anthropic.

Taken together, the two stories give you a concrete read on the compute buildout sitting behind Claude Code's roadmap. And looking ahead, Bloomberg is also reporting that Anthropic plans a wide release of its Mythos-class models in the coming weeks, alongside this week's Opus four point eight launch. So the model menu you see today may shift again soon. That's it for the brief.

I'm Shannon, and we'll see you next week.