Claude Code 2.1.77 shipped today — and the headline feature is one many have been waiting for: default output tokens have been doubled to 64K, with a new upper bound of 128K.
What Does This Mean in Practice?
If you use Claude Code to generate longer files, run complex refactors, or create entire components at once, you’ve probably hit truncated responses regularly. The 32K limit was simply too tight for many workflows. With 64K as the new default — and the ability to go up to 128K — that should happen a lot less.
Other Improvements
Beyond the output upgrade, 2.1.77 brings some solid quality-of-life improvements:
A new allowRead sandbox setting enables read access within restricted regions. This matters for teams working with strict sandbox rules who still need certain directories to be readable.
macOS startup is now roughly 60ms faster — sounds small, but it adds up when you’re launching Claude Code multiple times a day. Session resuming has also been accelerated, so you can jump back into running sessions more quickly.
Additionally, bugs in bash command handling, terminal UI rendering, and clipboard operations in environments like tmux and iTerm2 have been fixed.
My Take
2.1.77 isn’t a feature fireworks show — it’s a solid quality-of-life update. But the output token doubling is genuinely important. It removes one of the most common frustrations in daily workflows. Combined with the 1M context window from 2.1.75 and MCP Elicitation from 2.1.76, Claude Code has received three significant upgrades in a single week.
If you’re still on an older version: now’s a good time to update.
Sources: