An update on recent Claude Code quality reports
Anthropic investigated user reports of degraded performance in Claude Code, the Agent SDK, and Cowork, and traced them to three separate changes: a lowered default reasoning effort, a bug that cleared session context on every turn after idle, and a prompt tweak to limit verbosity. All issues were reverted or fixed by April 20 (v2.1.116), and Anthropic is adding stricter testing, broader internal dogfooding, and more targeted evals to prevent similar regressions.
- Three unrelated changes (lowered reasoning effort, a context-clearing bug, and a verbosity-limiting prompt tweak) independently degraded Claude Code quality, and all were rolled back or fixed by April 20 (v2.1.116).
- A March 26 bug meant to prune old reasoning after an hour of idle time instead cleared context on every turn, causing Claude to forget logic, repeat itself, and make bad tool calls—while also burning through users' rate limits faster via cache misses.
- The default reasoning effort was dropped from high to medium on March 4 to reduce latency, but users found it dumbed-down, so it was reverted on April 7 to x-high (Opus 4.7) or high (other models).
- Because the three issues hit different traffic subsets at different times, users experienced inconsistent symptoms—slowdowns, forgetfulness, quality drops—making the problem hard to diagnose until late April.
Over the past month, Claude users reported that coding responses felt weaker and more error-prone. Anthropic traced the problem to three separate releases affecting Claude Code and the Claude Agent SDK—none touched the API itself—and fixed them all by April 20 with version 2.1.116. First, on March 4 they switched the default reasoning effort from high to medium to curb long pauses, only to hear that medium felt dumbed-down. They rolled that back on April 7, restoring x-high effort for Opus 4.7 and high effort for other models.
Next came a March 26 change meant to speed up idle sessions by pruning old reasoning from context after an hour. A bug kept clearing history on every turn, so Claude forgot its own logic, repeated itself and made odd tool calls. That one finally got patched on April 10 in v2.1.101. The repeated cache misses also drained users’ rate limits faster, which added to the confusion.
Finally, on April 16 they introduced a new system prompt to tone down verbosity in Opus 4.7. Combined with earlier prompt tweaks, it cut into coding quality. They reverted that change on April 20 as well. Since each tweak hit different subsets of traffic at different times, users saw a patchwork of slowdowns, forgetfulness, and drop-offs in output quality until late April, when all three problems were resolved.