fix(proxy_server): populate model_info when cost keys are present but None (#40741)

[tusharjamunkar]

TLDR

Problem this solves:

• ModelInfo serialization dumps fields with None values
• proxy_server.py checked if k not in model_info:
• Existing None values blocked model cost map enrichment
• Missing pricing broke concurrent budget reservation

How it solves it:

• Check if model_info.get(k) is None: at both merge sites
• Missing and None fields get enriched with litellm cost data
• Preserves custom user-specified pricing values

User Flow

Before: a proxy user configures a model deployment without explicit per-token pricing, and concurrent budget reservations fail to estimate spend, allowing requests through with $0 estimated spend

1. Proxy initializes a model deployment with default ModelInfo fields (where input_cost_per_token=None, etc.)
2. proxy_server._enrich_model_info_with_litellm_data sees k in model_info (with value None) and skips filling in litellm's default pricing
3. The model's cost map remains empty/None, leading concurrent request budget reservations to evaluate spend at $0.0

After: the proxy initializes the model deployment, enriches unset/None fields with standard model pricing, and preserves any custom pricing overrides

1. Proxy initializes the same model deployment with default ModelInfo
2. proxy_server._enrich_model_info_with_litellm_data checks if model_info.get(k) is None: and populates input_cost_per_token, output_cost_per_token, and other metadata from the default cost map
3. The model has accurate pricing info, enabling concurrent budget reservations to accurately track spend

Relevant issues

Fixes #40741

Linear ticket

Pre-Submission checklist

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

• I have added meaningful tests
• The handful of test files covering my change pass locally, e.g. uv run pytest tests/test_litellm/<your_test_file>.py -v. Leave the suites (make test-unit-*, make test-unit) to CI: it finishes in ~15 minutes where a laptop takes an hour or more
• My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

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

Screenshots / Proof of Fix

Before (dfab479)

1. Run reproduction script querying _enrich_model_info_with_litellm_data with default ProxyModelInfo
2. Observed output: input_cost_per_token: None, max_tokens: None, mode: None (pricing was skipped because keys were present with None value)

After (80f4d7e)

1. Run reproduction script querying _enrich_model_info_with_litellm_data with default ProxyModelInfo
2. Observed output: input_cost_per_token: 5.5e-06, max_tokens: 128000, mode: 'chat' (pricing and token limits populated from cost map)
3. Run with custom override: input_cost_per_token: 0.00099 is preserved

Type

🐛 Bug Fix
✅ Test

Caveats (if any)

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

[greptile-apps]

Greptile Summary

This PR updates both model-info enrichment paths to populate metadata whose existing value is None, while preserving explicit user values.

• Populates pricing, token limits, and other metadata from LiteLLM’s model-cost data when unset.
• Adds direct regression coverage for both enrichment functions.
• Verifies that custom pricing overrides remain unchanged.

Confidence Score: 5/5

The PR appears safe to merge, with focused regression coverage for both affected enrichment paths.

No actionable new issues remain. The previously reported test gap is resolved by independently invoking _get_proxy_model_info with a fresh model whose metadata fields are None.

Important Files Changed

Filename	Overview
litellm/proxy/proxy_server.py	Both model-info merge sites now treat missing and None values as unset while preserving non-None overrides.
tests/test_litellm/proxy/proxy_server/test_routes_model_info.py	Adds independent regression tests for both merge paths and coverage for preserving custom pricing.

_{Reviews (2): Last reviewed commit: "test(proxy_server): add independent regr..." | Re-trigger Greptile}

[codspeed-hq]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing tusharjamunkar:fix/proxy-enrich-model-info-none (80f4d7e) with litellm_internal_staging (dfab479)}

[tusharjamunkar]

@greptileai

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!