Skip to content

Add au.anthropic.claude-opus-4-8 to bedrock model config and pricing - #1027

Merged
b4s36t4 merged 1 commit into
mainfrom
fix/bedrock-au-opus-4-8-config
Nov 12, 2026
Merged

b4s36t4 merged 1 commit into
mainfrom
fix/bedrock-au-opus-4-8-config

Conversation

@itsthusharashenoi

Copy link
Copy Markdown
Contributor

Pylon: https://app.usepylon.com/issues?issueNumber=9901&nid=56d9e34d-d1f5-4ed5-b43a-c3f58f44d370&nsrc=slack&ntype=issue_sla_breached

What is the issue

The au.anthropic.claude-opus-4-8 Bedrock inference profile had no entry in the model config, so it fell back to the Bedrock default block. The UI therefore showed Max Tokens as 2048 (default 512) instead of Opus 4.8's real limit of 128K output tokens (customer ticket #9901). The pricing file had no entry for this profile either.

What is the fix

  • general/bedrock.json: added au.anthropic.claude-opus-4-8 with max_tokens maxValue: 128000. The thinking, tool_choice, type and removeParams settings are copied from the existing us./eu. Opus 4.8 entries.
  • pricing/bedrock.json: added pay-as-you-go and batch pricing for the profile. The prices match the us./eu. geo profiles (10% premium over the base model). This is consistent with how au.anthropic.claude-opus-5-5 is priced against its base.

Max output (128K) is per the Anthropic and AWS Bedrock docs for Opus 4.8.

Not included (follow-ups):

  • jp.anthropic.claude-opus-4-8 is also missing from both files.
  • albus/src/configs/model-configs/general/bedrock.json has the same gap.
  • The Opus 4.8 thinking schema (enabled / budget_tokens) differs from the AWS docs, which say adaptive only.

How to test

  1. Run python3 scripts/check_duplicate_keys.py. It should report no duplicate keys.
  2. Run node -e "const j=require('./general/bedrock.json');console.log(j['au.anthropic.claude-opus-4-8'].params[0])". It should print { key: 'max_tokens', maxValue: 128000 }.
  3. Run node -e "const j=require('./pricing/bedrock.json');console.log(j['au.anthropic.claude-opus-4-8'].pricing_config.pay_as_you_go.request_token)". It should print { price: 0.00055 }.
  4. After the config is deployed, open the model in the UI (provider bedrock-au, model au.anthropic.claude-opus-4-8). Max Tokens should show 128000 rather than 2048.

Co-authored-by: Cursor <cursoragent@cursor.com>
Copilot AI balanced review requested due to automatic review settings November 11, 2026 08:51

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟢 Approval recommended

No active review issues are recorded for the Australian Bedrock profile configuration and pricing changes.

No open issues 🎉

What changed in this PR

Adds explicit configuration and pricing for the Australian Bedrock profile au.anthropic.claude-opus-4-8, replacing the generic fallback responsible for the incorrect Max Tokens limit in the UI.

Changes:

  • Sets the maximum output to 128,000 tokens and mirrors the existing US/EU profile settings.
  • Adds pay-as-you-go, cache, and batch pricing matching the US/EU geographic profiles.
File Description
pricing/​bedrock.json Adds token, cache, and batch rates for the Australian profile.
general/​bedrock.json Adds the profile’s token limit, parameters, and supported capabilities.

Review effort: Balanced

@b4s36t4
b4s36t4 merged commit c7b403c into main Nov 12, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants