Claude Code 2.1.293 adds Haiku 5.5 and closes a memory leak on HTTP MCP
In /model the effort dial wrapped past the highest level with the arrow keys — and could quietly save Low as a model's default.
Claude Haiku 5.5: $0.10 per million input tokens, 1M context, and a first effort dial
Haiku 4.5 scored a flat 0.0% on Terminal-Bench 4.0. The 5.5 gets 39.2 — at a tenth of the price.
GPT-6 and Intelligent UI for everyone: ChatGPT now answers with interfaces
Behind the generated interfaces sits a compiler that renders the answer while the model is still writing it. Codex and Work are left out.
Cisco and Carnegie Mellon study: 8 of 13 models recommend pricier options to wealthy users
Strip out the financial profile and let the models read only an email inbox, and much of the gap survives. Gemini read the two finance emails first in 97% of trials.
Managed Agents: web_fetch only retrieves URLs that already appeared in the session
A URL from Claude's own output doesn't count. Neither does one from the system prompt, an attached document, or the output of bash or an MCP tool.
Max and Team plans get monthly API credits, Sonnet 5.5 halves its cache read price
Max 20x brings $200 a month, a Team plan up to $500 — pooled across every user. You have to claim it yourself, through a linked Console organization.
Claude Code 2.1.290 takes back eight things auto-approval used to let through
A mod you downloaded could unload your organization's plugin. Now the mod is the one that goes.
Claude as a diary: Anthropic reported an entry to police, and Florida is charging under Statute 836.10
Anthropic may hand over user data in narrow emergencies, when it believes that prevents death or serious injury. That's the clause it pulled here.