For the past 8 hours or so, many of our queries with GPT-OSS have been timing out or returning long strings of “\n\n\n\n\n\n\n\n\n\n\n\n” up to the token limit. I’m pretty sure this is not anything on our end - we have tried reverting to older commits that we know are stable and still observe the same issue.
Other models seem to work fine - it’s only GPT-OSS that’s affected.
Could you post your full cURL and response object?
I’m trying this and it runs fine. Are you using tools?
{
"model": "openai/gpt-oss-120b",
"messages": [
{
"role": "user",
"content": "Tell me a fun fact about the moon in one sentence."
}
]
}
Update, this also seems to work right now:
{
"model": "openai/gpt-oss-120b",
"messages": [
{
"role": "user",
"content": "[ALWAYS USE A TOOL] What is the weather in DC?"
}
],
"tools": [
{
"type": "function",
"function": {
"name": "get_current_weather",
"description": "Get the current weather in a given location",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "The city and state, e.g. San Francisco, CA"
},
"unit": {
"type": "string",
"enum": [
"celsius",
"fahrenheit"
]
}
},
"required": [
"location"
]
}
}
},
{
"type": "function",
"function": {
"name": "catch_all",
"description": "A general-purpose tool to handle requests that don't match any other available tools. Use this when no other tool is applicable or when the task is too complex or open-ended for a specific tool.",
"parameters": {
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "The user's query or request that needs to be handled"
}
},
"required": [
"query"
]
}
}
}
],
"tool_choice": "auto"
}
It’s been very inconsistent all day. Some times it works perfectly fine, other times it behaves extremely erratically and fails over and over. I’ll see if I can find a concrete example the next time it fails. Most of the times it has occurred thus far have been in the context of our multi-agent workflows with tool calling.
yes please, I’d love to see your logs / the API responses from the erroring calls
I can confirm this as well. Padding a whole bunch of repetitive stuff at the end of tool calls. GPT120B model for me. Not seeing it now but please investigate!
Thank you for the reports all, we’re investigating.
Can confirm, facing the same issue, with less than 10k tokens being sent to GPT OSS with medium/high reasoning, I am still experiencing chat message limit exceeded, and I’m pretty sure it’s not generating 100k tokens worth of valid reasoning, it’s doing some sort of gibberish and then crashing at the end.
Thanks so much for reporting, the errors should be fixed now.
(We’ve been adding some new improvements to the GPT OSS models that we’ll be able to announce shortly!)