add us gov models

[mubashir1osmani]

Relevant issues

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Type

🆕 New Feature
🐛 Bug Fix
🧹 Refactoring
📖 Documentation
🚄 Infrastructure
✅ Test

Changes

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Mar 27, 2026 8:12pm

[codspeed-hq]

Merging this PR will not alter performance

✅ 16 untouched benchmarks

---

_{Comparing mubashir1osmani:aws-gov-models (d8a2a3f) with main (d949085)}

[greptile-apps]

Greptile Summary

This PR adds AWS GovCloud (us-gov-east-1 / us-gov-west-1) support for Claude Sonnet 4.5 (claude-sonnet-4-5-20250929-v1:0) by adding three new JSON entries and correcting two pre-existing ones. The changes land in both the root model_prices_and_context_window.json and its backup copy in litellm/.

Key changes:

• Adds bedrock/us-gov-east-1/anthropic.claude-sonnet-4-5-20250929-v1:0 and bedrock/us-gov-west-1/anthropic.claude-sonnet-4-5-20250929-v1:0 (bedrock direct-API entries, max_output_tokens: 8192, consistent with all other GovCloud bedrock direct-API entries in the file)
• Adds us-gov.anthropic.claude-sonnet-4-5-20250929-v1:0 (bedrock_converse cross-region inference profile, max_output_tokens: 64000) — the us-gov prefix is already recognised by get_bedrock_cross_region_inference_regions() in common_utils.py, so routing will work correctly
• Bumps max_output_tokens / max_tokens from 4096 → 8192 on the two pre-existing bedrock/us-gov-*/claude-sonnet-4-5-20250929-v1:0 entries, aligning them with the new anthropic.-prefixed siblings
• The new bedrock_converse entry still omits search_context_cost_per_query, which all equivalent cross-region entries (us., au.) include — this was flagged in a prior review thread and remains unresolved

Process note: The PR checklist items for tests and CI runs are unchecked. While a pure JSON config addition is low-risk, the project requires at least one new unit test per contribution.

Confidence Score: 4/5

Safe to merge for routing correctness; one cost-estimation field remains missing from the new bedrock_converse entry per a prior thread, and no tests were added

The routing infrastructure already supports the us-gov. prefix, all new entry costs mirror the equivalent us. entry, and the max_output_tokens inconsistency flagged in the prior thread was resolved by updating the pre-existing entries. Score stops at 4 because the search_context_cost_per_query omission from the new us-gov.anthropic.claude-sonnet-4-5-20250929-v1:0 entry (called out in a previous review cycle) is still unaddressed, and the mandatory test requirement is unfulfilled.

model_prices_and_context_window.json — specifically the us-gov.anthropic.claude-sonnet-4-5-20250929-v1:0 entry around line 28463

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Adds two new bedrock direct-API entries for us-gov-east-1 and us-gov-west-1 (with anthropic. prefix), one new bedrock_converse cross-region inference profile entry (us-gov.), and updates existing claude-sonnet-4-5-20250929-v1:0 GovCloud bedrock entries from 4096 → 8192 max output tokens; the new bedrock_converse entry is missing search_context_cost_per_query (see prior thread)
litellm/model_prices_and_context_window_backup.json	Mirror of all changes in the root JSON — identical set of additions and max_output_tokens bumps for GovCloud Sonnet 4.5 entries

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[LiteLLM request] --> B{Parse model prefix}
    B -->|bedrock/us-gov-east-1/...| C[bedrock direct API\nus-gov-east-1]
    B -->|bedrock/us-gov-west-1/...| D[bedrock direct API\nus-gov-west-1]
    B -->|us-gov.*| E{get_bedrock_cross_region_inference_regions\ncontains 'us-gov'?}
    E -->|Yes| F[bedrock_converse\ncross-region inference profile\nroute to us-gov-east-1 or us-gov-west-1]
    E -->|No| G[Error: unrecognised prefix]
    C --> H[anthropic.claude-sonnet-4-5-20250929-v1:0\nmax_output_tokens: 8192]
    D --> I[anthropic.claude-sonnet-4-5-20250929-v1:0\nmax_output_tokens: 8192]
    F --> J[anthropic.claude-sonnet-4-5-20250929-v1:0\nmax_output_tokens: 64000]

Loading

_{Reviews (2): Last reviewed commit: "added max tokens" | Re-trigger Greptile}

[greptile-apps]

max_output_tokens inconsistency with pre-existing sibling entry

The new bedrock/us-gov-east-1/anthropic.claude-sonnet-4-5-20250929-v1:0 entry uses max_output_tokens: 8192 and max_tokens: 8192, but the pre-existing entry for the same model in the same region — bedrock/us-gov-east-1/claude-sonnet-4-5-20250929-v1:0 (lines 7669–7688) — uses max_output_tokens: 4096 and max_tokens: 4096.

These two keys map to the same underlying model (Bedrock model IDs with and without the anthropic. prefix are interchangeable). Having different max_output_tokens for the same model creates confusion and potential bugs: a user who looks up context window information will get different results depending on which key format they happen to use.

The same inconsistency exists for the west-1 variant (bedrock/us-gov-west-1/anthropic.claude-sonnet-4-5-20250929-v1:0 reports 8192 while bedrock/us-gov-west-1/claude-sonnet-4-5-20250929-v1:0 reports 4096).

For reference, other bedrock/ format Claude models in these us-gov regions consistently use max_output_tokens: 8192 (e.g. bedrock/us-gov-west-1/anthropic.claude-3-7-sonnet-20250219-v1:0 and bedrock/us-gov-west-1/anthropic.claude-3-5-sonnet-20240620-v1:0), suggesting 8192 is correct and the pre-existing 4096 entries should be updated to match.

[greptile-apps]

Missing search_context_cost_per_query compared to equivalent cross-region entries

All other bedrock_converse cross-region Claude Sonnet 4.5 entries include a search_context_cost_per_query block:

• us.anthropic.claude-sonnet-4-5-20250929-v1:0 — has search_context_cost_per_query
• au.anthropic.claude-sonnet-4-5-20250929-v1:0 — has search_context_cost_per_query

The new us-gov.anthropic.claude-sonnet-4-5-20250929-v1:0 entry omits this field entirely. If the web search tool is not available on AWS GovCloud, this is intentional and should be noted (a comment or doc reference would help). If it is available, the field should be added for cost-estimation correctness.