#Logprobs parameter support issue

4 messages · Page 1 of 1 (latest)

raven smeltBOT
#

I'm an AI assistant - my answers may be incorrect and I cannot see screenshots. Please verify information with the OpenRouter docs.

To ensure that your requests are always routed to a provider that supports the logprobs parameter, you need to:

  1. Check the Supported Parameters for each provider on the model's page to confirm logprobs support.
  2. Use the API's provider selection features to specify a provider that supports logprobs.

For example, for Llama 3.1 70B Instruct, you can see which providers support logprobs by looking for "Logprobs" under "Supported parameters" in the provider table. Providers like Lambda, Nebius AI Studio, and Hyperbolic currently list "Logprobs" as supported Llama 3.1 70B Instruct Providers.

You can specify the provider in your API request by using the provider-specific route:

POST https://openrouter.ai/api/v1/chat/completions
{
  "model": "meta-llama/llama-3.1-70b-instruct",
  "provider": "lambda", // or "hyperbolic", etc.
  "logprobs": true,
  ...
}

"According to the OpenRouter model catalog, you can view models and their available providers, including supported parameters like logprobs."

There is no automated way to force fallback only to providers supporting logprobs; you must explicitly specify a provider that lists logprobs as a supported parameter Llama 3.1 70B Instruct Providers.

I recommend always checking the OpenRouter Models page for the latest provider capabilities, as support for parameters can change.

-# This is an automated response powered by Answer HQ. Use !mute to stop the bot from responding.

The author you are looking for could not be found.

The model you are looking for could not be found.

The author you are looking for could not be found.

OpenRouter

Browse models on OpenRouter

primal wagon
#

I want an auto solution. I don't want to have to manually select a new provider every time one of them suddenly stops supporting the logprobs parameter.

raven smeltBOT
#

There is no fully automatic solution to guarantee that only providers supporting logprobs will be used for every request. OpenRouter’s routing system sends requests to providers “that are able to handle your prompt size and parameters, with fallbacks to maximize uptime” Llama 3.1 405B Instruct Providers. However, if a provider suddenly drops logprobs support, OpenRouter may fall back to another provider that does not support it, resulting in errors or missing data.

Key points from the documentation:

  • “OpenRouter routes requests to the best providers that are able to handle your prompt size and parameters, with fallbacks to maximize uptime.”
  • Supported parameters (like logprobs) are listed per provider, but routing is not parameter-guaranteed unless you restrict your request to specific providers.

There is currently NO way to have OpenRouter automatically guarantee logprobs support across all fallback providers. The only way to ensure every response includes logprobs is to specify a provider that supports it.
If provider support changes, you must update your configuration to select a supported provider Llama 3.1 405B Instruct Providers.

If you need this feature, you may want to request it as a feature enhancement or check for updates with the mods or support. For now, review supported parameters on the OpenRouter Models page before sending requests.

"OpenRouter routes requests to the best providers that are able to handle your prompt size and parameters, with fallbacks to maximize uptime." Llama 3.1 405B Instruct Providers

*"When an error occurs in ...

The model you are looking for could not be found.

OpenRouter

Browse models on OpenRouter

The highly anticipated 400B class of Llama3 is here! Clocking in at 128k context with impressive eval scores, the Meta AI team continues to push the frontier of open-source LLMs.

Meta's latest class of model (Llama 3. Run Llama 3.1 405B Instruct with API

#

...an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it."* GLM 4.5 Air Providers

There is currently no way to force fallback only to providers that support specific parameters like logprobs.

-# This is an automated response powered by Answer HQ. Use !mute to stop the bot from responding.

GLM-4.5-Air is the lightweight variant of our latest flagship model family, also purpose-built for agent-centric applications. Run GLM 4.5 Air (free) with API