[BUG] No small/fast model in APAC
Environment
- Platform (select one):
- [ ] Anthropic API
- [x] AWS Bedrock
- [ ] Google Vertex AI
- [ ] Other: <!-- specify -->
- Claude CLI version: 1.0.102
- Operating System: ubuntu 20.04
- Terminal: vscode embedded terminal
Bug Description
It appears that claude code will not use the small/fast model if the region is an APAC one
Steps to Reproduce
CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=ap-southeast-2 ANTHROPIC_MODEL=apac.anthropic.claude-sonnet-4-20250514-v1:0 ANTHROPIC_SMALL_FAST_MODEL_REGION=ap-southeast-2 ANTHROPIC_SMALL_FAST_MODEL=apac.anthropic.claude-3-haiku-20240307-v1:0 claude- "Is my README.md up to date?"
- /cost
Expected Behavior
I expected to see some usage by model for claude-sonnet, and some for haiku
Actual Behavior
I only see usage for claude-sonnet
Additional Context
If I use CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=us-west-2 claude, I do get a split in usage by model.
I also get a split in usage if I use CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=us-west-2 ANTHROPIC_MODEL=us.anthropic.claude-sonnet-4-20250514-v1:0 ANTHROPIC_SMALL_FAST_MODEL_REGION=us-west-2 ANTHROPIC_SMALL_FAST_MODEL=us.anthropic.claude-3-haiku-20240307-v1:0 claude
I cannot use CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=ap-southeast-2 claude, as claude code assumes the model starts with "us.", when in that region it starts with "apac.".
I have confirmed I have access to all the models being used, both in Model Access and IAM.
7 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
These are all different issues from this.
Just to confirm availability and IAMs, can you try setting that Haiku model as your main model?
CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=ap-southeast-2 ANTHROPIC_MODEL=apac.anthropic.claude-3-haiku-20240307-v1:0 claude@ant-kurt
API Error: 400 messages.0.content.2.text.cache_control: Extra inputs are not permittedSo I've tried the next most logical thing...
CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=ap-southeast-2 ANTHROPIC_MODEL=apac.anthropic.claude-3-haiku-20240307-v1:0 DISABLE_PROMPT_CACHING=1 claude- which works and reports the usage as "claude-3"CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=us-west-2 ANTHROPIC_MODEL=us.anthropic.claude-3-haiku-20240307-v1:0 claudesaysAPI Error: 400 You invoked an unsupported model or your request did not allow prompt caching. See the documentation for more information.CLAUDE_CODE_USE_BEDROCK=1 AWS_REGION=us-west-2 ANTHROPIC_MODEL=us.anthropic.claude-3-haiku-20240307-v1:0 DISABLE_PROMPT_CACHING=1 claudeworks, and reports usage for "claude-3-5-haiku" and "claude-3"We're looking into this, thanks for sharing the above info.
For the time being, you'll see a few degradations (around summarization and bash tool). You can mitigate this by either using a region with Claude 3.5 Haiku, or setting
DISABLE_PROMPT_CACHING=1(which will disable for the main model as well, impacting performance and cost).v1.0.113 added a few env vars that should resolve this for you -
DISABLE_PROMPT_CACHING_HAIKU,DISABLE_PROMPT_CACHING_SONNET,DISABLE_PROMPT_CACHING_OPUS. This allows you to keep prompt caching enabled for your main loop model, while disabling for your unsupported Haiku version specifically.This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.