Where does the bug appear (feature/product)?
Cursor IDE
Describe the Bug
Since around the 24th of Oct 2025, Cursor’s Bedrock integration for Claude Sonnet 4.5 automatically includes the following field in API calls:
"anthropic_beta": ["context-1m-2025-08-07"]
This flag explicitly enables the 1-million token context mode on AWS Bedrock.
That mode:
- Doubles token pricing ($0.003 → $0.006 input, $0.015 → $0.0225 output per 1K tokens)
- Uses a separate, much lower quota pool (preview feature with stricter rate limits)
- Is not user-configurable or documented in Cursor settings
Example CloudWatch log extract:
"thinking": {"type": "enabled", "budget_tokens": 6000},
"anthropic_beta": ["context-1m-2025-08-07"]
Even small requests (<200K tokens) are billed and throttled under the long-context tier.
This appears to have been introduced recently, as older requests used the 200K mode without the beta flag.
While having access to the 1M context window is a much welcome feature, it should be on-demand and invoked only with user consent.
Enabling it automatically makes the product effectively unusable for AWS Bedrock users due to the higher cost and much stricter service quotas associated with the long-context mode.
Requests:
- Confirm whether this behavior is intentional.
- Add a setting or environment variable to disable long-context mode.
- Make the 1M context window available only on user opt-in, not by default.
Steps to Reproduce
-
Open Cursor IDE and select Claude Sonnet 4.5 as the model (via AWS Bedrock key).
-
Send any prompt - it can be small, e.g. 1+1= (see screenshot)
-
Check your AWS CloudWatch logs for the corresponding
InvokeModelWithResponseStreamevent. -
Observe that the request payload includes:
"anthropic_beta": ["context-1m-2025-08-07"] -
Note that the same request is billed under Claude Sonnet 4.5 – Long Context pricing and is subject to stricter rate limits, even though the input is below 200K tokens.
Expected Behavior
Cursor should default to 200K context mode unless the user explicitly choses to enable the 1M context mode.
Screenshots / Screen Recordings
Operating System
MacOS
Current Cursor Version (Menu → About Cursor → Copy)
Version: 1.7.54
VSCode Version: 1.99.3
Commit: 5c17eb2968a37f66bc6662f48d6356a100b67be0
Date: 2025-10-21T19:07:38.476Z
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Darwin arm64 24.5.0
For AI issues: which model did you use?
us.anthropic.claude-sonnet-4-5-20250929-v1:0
global.anthropic.claude-sonnet-4-5-20250929-v1:0
For AI issues: add Request ID with privacy disabled
Request ID: 8d740de6-b688-4c4a-b652-1769519fa073
Error:
ERROR_OPENAI_RATE_LIMIT_EXCEEDED
Hit your Anthropic API key rate limit.
You're using your own API key, and we received a rate limit error from Anthropic.
Severity: High - this behavior is leading to substantial unexpected billing (hundreds of dollars per day) and makes Cursor unusable under AWS Bedrock quotas for the 1M context model.
Does this stop you from using Cursor
Yes - Cursor is unusable

