feat(openrouter): add stealth/union-alpha to the model cost map - #41576

TLDR

Problem this solves:

• OpenRouter's new stealth model stealth/union-alpha is missing from the cost map
• Calls work today but cost tracking and capability lookups have no entry for it

How it solves it:

• Adds openrouter/stealth/union-alpha to the cost map and its backup copy
• Values come from GET https://openrouter.ai/api/v1/models: 262,144 context, 131,072 max completion tokens, $0 in/out, text+image input, tools, tool_choice, response_format

User Flow

Before: a developer routing to Union Alpha through the gateway gets a working answer but the model is unknown to the gateway

1. The proxy admin adds a deployment with model: openrouter/stealth/union-alpha and an OPENROUTER_API_KEY
2. The developer sends POST https://litellm-domain/v1/chat/completions with "model": "union-alpha" and gets a 200 with a normal completion
3. GET https://litellm-domain/model/info shows the deployment with max_input_tokens: null, supports_vision: null and supports_function_calling: null, and SDK users calling completion_cost on the response get This model isn't mapped yet

After: the same request is priced and described

1. The proxy admin adds the same deployment
2. The developer sends the same POST https://litellm-domain/v1/chat/completions and gets the same 200
3. GET https://litellm-domain/model/info now reports max_input_tokens: 262144, max_output_tokens: 131072, supports_vision: true, supports_function_calling: true and $0 pricing, and completion_cost returns 0.0

Relevant issues

Affected release

Linear ticket

Resolves LIT-7988

Closes LIT-7988

Pre-Submission checklist

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

• I have added meaningful tests (data-only change, covered by the existing cost map schema 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 (no test file touched; verified via live proxy below)
• 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

Live calls against OpenRouter with a real key, through both the SDK and a running proxy. Before runs on a worktree at the merge base (port 4001), After runs on the PR tip (port 4000), both started with LITELLM_LOCAL_MODEL_COST_MAP=True python litellm/proxy/proxy_cli.py --config config.yaml where config.yaml has one deployment, model_name: union-alpha -> model: openrouter/stealth/union-alpha

Source row, fetched with curl -s https://openrouter.ai/api/v1/models on 2026-09-17:

id: stealth/union-alpha
context_length: 262144
top_provider.max_completion_tokens: 131072
pricing: prompt 0, completion 0
architecture.input_modalities: text, image
supported_parameters: max_tokens, response_format, temperature, tool_choice, tools, top_p

Commands (run once per port):

curl -s http://localhost:$PORT/chat/completions -H "Authorization: Bearer sk-1234" -H "Content-Type: application/json" \
  -d '{"model":"union-alpha","messages":[{"role":"user","content":"Reply with exactly: union alpha ok"}],"max_tokens":20}'
curl -s http://localhost:$PORT/model/info -H "Authorization: Bearer sk-1234"

Before (merge base, port 4001)

{"id":"gen-1789629942-sBM8JGOMapIjGxBOj26f","model":"union-alpha","choices":[{"finish_reason":"stop","message":{"content":"union alpha ok","role":"assistant"}}],"usage":{"completion_tokens":7,"prompt_tokens":6,"total_tokens":13},"provider":"Stealth"}

model_info: {'max_input_tokens': None, 'max_output_tokens': None, 'input_cost_per_token': 0, 'output_cost_per_token': 0, 'supports_vision': None, 'supports_function_calling': None, 'supports_response_schema': None}

SDK at the merge base, litellm.completion(model='openrouter/stealth/union-alpha', ...) returns union alpha ok but litellm.completion_cost(completion_response=r) raises ValueError: This model isn't mapped yet. Add it here - https://github.com/BerriAI/litellm/blob/main/model_prices_and_context_window.json

After (3d0fd12, port 4000)

{"id":"gen-1789630010-JMbKbJjhd4Hy2ykz8DZV","model":"union-alpha","choices":[{"finish_reason":"stop","message":{"content":"union alpha ok","role":"assistant"}}],"usage":{"completion_tokens":7,"prompt_tokens":24,"total_tokens":31},"provider":"Stealth"}

model_info: {'max_input_tokens': 262144, 'max_output_tokens': 131072, 'input_cost_per_token': 0, 'output_cost_per_token': 0, 'supports_vision': True, 'supports_function_calling': True, 'supports_response_schema': True}

SDK at the PR tip, same litellm.completion call returns union alpha ok with usage 24/7 and litellm.completion_cost(completion_response=r) returns 0.0

Type

New Feature

Caveats (if any)

Low

• Stealth model: OpenRouter may rename or remove it after the preview
• Pricing is $0 during the preview; a later paid price will need a follow-up

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

Link to Devin session: https://app.devin.ai/sessions/1bb1affe356b4dd795450e529e2cff25
Open in Devin Desktop: https://app.devin.ai/desktop/session/1bb1affe356b4dd795450e529e2cff25?variant=devin