Sign in Start free

Solved

Why was my model request rejected?

ToRun Team · Replies: 1 · Views: 11

I've been using ToRun for a few weeks and overall it's been great, but today I ran into a confusing error. I was in the middle of a chat session — I had uploaded an image and asked the model a follow-up question — and instead of a response I got something like "request could not be fulfilled." No detail beyond that.

I'm not sure where to start debugging this. I checked my wallet and it still shows a small balance, so I don't think it's a funds issue, but I could be wrong. I also noticed the mode I was using had automatically selected a model I don't normally see in that slot. I hadn't changed any settings.

A few questions I'm hoping someone can clarify:

  1. Can a model be selected that technically doesn't support something I asked for, like vision? How would I know?
  2. Is there a case where my balance looks non-zero but the request still fails for funding reasons?
  3. Could the image push my context over some limit?
  4. Is there a way to tell whether ToRun tried to fall back to another provider and that also failed?
  5. Could a content moderation policy have blocked it silently?

I want to understand all the possible causes so I can figure out which one actually hit me here. What are the things I should check?

1 Replies

ToRun Team ·14 Jun 2026 Solved

Good questions — there are five distinct causes for a rejected request, and they can look identical from the outside. Here is what to check in order.

1. Capability mismatch. The routing pipeline selects a model based on the capabilities your request needs. If the chosen model does not support vision (or whichever capability you triggered), the request is rejected before it reaches the provider. Check the model shown in the session header and verify it has the Vision capability listed under its details. Switching modes or explicitly choosing a vision-capable model in the model selector will fix this.

2. Funding gap. A non-zero wallet balance can still be insufficient if the estimated cost for that request (image input, large context) exceeds what remains. Subscriptions also have a period allowance that can be exhausted independently of your wallet. Look at both the wallet balance and any active subscription allowance on your billing page.

3. Context window exceeded. An uploaded image counts toward the token budget. If the combined input (system prompt + history + image tokens + your message) exceeds the model's context window, the request is rejected. The context indicator in the chat bar shows current usage. Clearing history or switching to a model with a larger window resolves this.

4. Provider outage with exhausted fallback. ToRun tries fallback providers automatically, but if every provider in the fallback chain for that capability set is unreachable, the request fails. You can check provider status on the status page.

5. Moderation block. If the input or the image triggered a moderation policy, the request is stopped before reaching the model. The session log will note a moderation event if this happened.

Check the session log for whichever signal appears — it will point to one of these five. If none is clear, open a support ticket and include the session ID.

Read-only on the public site
Open in the app