Hacktoberfest 2026: le issue che i maintainer hanno segnato per ottobre, aperte e adatte ai principianti. Sfoglia le issue Hacktoberfest

GPT-6 models are sent max_tokens and rejected (not treated as reasoning models)

Aperta
  • Qualcosa non va?
#1,605 1 commento 0 reazioni 1 assegnatario Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

@paullizer ci sta già lavorando.

Dal 2/10/2026.

Valutazione

Questa issue non è ancora stata valutata.

Descrizione

bug P1

Issue

GPT-6 deployments, such as gpt-6-astra, fail with a 400 whenever SimpleChat sets a response length:

Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead.

This affects:

  • every workflow model call once the model's token limits are configured, because the workflow budget always sets a response length;
  • chat when Response Length is set.

Found on a SimpleChat POC deployment running the React v2 branch.

Steps to Reproduce

  1. Add a GPT-6 deployment, such as gpt-6-astra, as a model and configure its token limits (context window, output limit, total-generation accounting).
  2. Run a workflow whose agent uses that model. Alternatively, chat with it with Response Length set.

Expected Behavior

SimpleChat sends max_completion_tokens and no temperature, as it does for GPT-5 and o-series models.

Actual Behavior

ModelEndpointBehavior.is_openai_reasoning_model in application/single_app/model_endpoint_clients.py recognizes only gpt-5, o1, o3 and o4 names (OPENAI_REASONING_MODEL_PREFIXES). GPT-6 deployments get max_tokens and a temperature, and the request is rejected. The same check exists on Development, main and the React v2 branch.

Impact

Workflows that use a GPT-6 model fail on their first model call. Prompt variables, content screening and orchestration share the same policy, so they send GPT-6 the same wrong parameters.

Notes

  • Fix: recognize gpt-6 like gpt-5, as a name prefix and as a family name inside deployment or display names. Add a regression test for the GPT-6 names and the workflow budget's parameter choice.
  • Already fixed on the React v2 branch in 0.261.219 (test_gpt6_reasoning_model_parameters.py, docs/explanation/fixes/GPT6_REASONING_MODEL_PARAMETERS_FIX.md). Needs porting to Development.
Lingua principale
Python
Stelle
157
Fork
117
Merge medio
4h 19m
PR unite (30g)
200

Preparare l'ambiente

Apri in Codespaces

Avvia il container di sviluppo del progetto nel browser, con il tuo account GitHub.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di microsoft/simplechat

Tutte le issue di microsoft/simplechat

Issue simili

Altre issue su Python

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.