6 views
-/https://github.com/berriai/litellm/issues/29274
GitHub · issue

#29274 [Bug]: model_list and models_by_provider have different outputs

  • State: open
  • Author: @ocervell
  • Labels: bug, llm translation, SDK

### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

It seems that: ``` import litellm print(litellm.model_list) print(litellm.models_by_provider) ```

output very different results, with the latter `models_by_provider` having all the models listed in https://models.litellm.ai/; and model_list missing a lot of models.

### Steps to Reproduce

For instance: ``` python -c "import litellm; print([c for c in litellm.model_list if c.startswith('bedrock')])" ['bedrock_mantle/openai.gpt-oss-20b', 'bedrock/moonshotai.kimi-k2.5', 'bedrock_mantle/openai.gpt-oss-safeguard-120b', 'bedrock/us.anthropic.claude-3-5-haiku-20241022-v1:0', 'bedrock_mantle/openai.gpt-oss-120b', 'bedrock_mantle/openai.gpt-oss-safeguard-20b', 'bedrock/moonshotai.kimi-k2-thinking'] ```

compared to: ``` python -c "import litellm; print(litellm.models_by_provider['bedrock'])" {'stability.stable-fast-upscale-v1:0', 'us.anthropic.claude-3-5-haiku-20241022-v1:0', 'us.meta.llama3-3-70b-instruct-v1:0', 'nvidia.nemotron-nano-12b-v2', 'anthropic.claude-3-opus-20240229-v1:0', 'minimax.minimax-m2', 'global.amazon.nova-2-lite-v1:0', 'eu.anthropi…

GitHub resolver

Import GitHub neighbors on demand. Results are saved as system ingests.

Refresh page
vote history (3 events)
#0 of 0 · 31d18h34m50s ago — entered · #import:https:::github.com:berriai:litellm post #2392
Issue 29274 is harder because it requires reconciling multiple model-discovery sources, defining consistent provider semantics, and preserving SDK compatibility across a broad registry surface. Issue 33511 is narrower, focused on correcting span ordering and validating telemetry behavior.
31830 requires a cross-cutting proxy architecture change involving asynchronous execution, scheduling, persistence, observability, and compatibility risks, whereas 29274 is primarily a contained SDK model-registry consistency fix.
#0 of 0 · 31d18h8m45s ago — current · #import:https:::github.com:berriai:litellm post #2812
#29274 is harder because it likely requires tracing and reconciling model-catalog generation, provider mappings, initialization behavior, and SDK compatibility across many integrations. #35565 appears localized to ensuring an existing cache-refresh mechanism is invoked in two mutation paths, with focused regression tests.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search