Claude Haiku 5.5 is priced at $0.10/M input and $0.50/M output tokens, or $0.50/M and $2.50/M above 100k input tokens. Claude Sonnet 5.5 cache reads are also now $0.10/M tokens, down from $0.20/M. Compare Haiku 5.5 on CursorBench:
@cursor_ai
-
Local AI agent keeps working without phone signal
By
–
The agent runs on your computer, so it keeps working even if your phone loses signal.
-
Cursor iOS app controls AI agents remotely
By
–
You can now control agents on your computer from your phone.
— Cursor (@cursor_ai) 6 octobre 2026
Check in, reply, or start new tasks from the Cursor iOS app. pic.twitter.com/11dcZhRyrxYou can now control agents on your computer from your phone. Check in, reply, or start new tasks from the Cursor iOS app.
-
SDK update for local TypeScript AI agents
By
–
Steering and system prompts work with local TypeScript agents. Background subagent results come back locally in TypeScript and Python. See everything new in the SDK:
-
Cursor lets users replace system prompt with custom rules
By
–
You can also replace Cursor's system prompt with your own. Rules, skills, and tool schemas still load. We're enabling it account by account.
-
MCP annotations let models distinguish read from destructive tool calls
By
–
Custom tools can now carry MCP annotations like readOnlyHint and destructiveHint, so the model can tell a lookup from a delete before it calls one.
-
Cursor SDK lets you steer running agents with run.steer()
By
–
You can now steer Cursor SDK agents while they run.
— Cursor (@cursor_ai) 5 octobre 2026
run.steer() adds your message to the next turn. If a subagent is mid-task, it moves to the background and keeps working. pic.twitter.com/1GHeV6kfKCYou can now steer Cursor SDK agents while they run. run.steer() adds your message to the next turn. If a subagent is mid-task, it moves to the background and keeps working.
-
Rollouts cloud agent auto-fixes regressions by opening issues
By
–
When Rollouts catches a regression, it finds the offending PR and opens an issue. One click starts a cloud agent to fix it. Rollouts usage credits are included through Oct 3.
-
GLM 5.3 and Flash now available in Cursor
By
–
GLM 5.3 and GLM 5.3 Flash are now available in Cursor! GLM 5.3 Max is the best-scoring open-weight model on CursorBench 4.0.
-
AI models hack benchmarks by retrieving solutions from internet
By
–
We're sharing new research on how models hack public benchmarks. The latest models, including Opus 4.8 and Composer 2.5, learn to retrieve solutions from the internet or git history. When we apply a stricter harness, eval scores drop significantly.