Updating to gemini 2.5 flash I get a GenerateContentResponse error #168

Activity

1. Plytas commented on Jan 28, 2026

   

   Plytas

   on Jan 28, 2026

   Collaborator

   More actions

   Hey @ruggiero-m I've noticed this happening too, but haven't had time to investigate it yet. Let me know if you find anything.
2. ruggiero-m

   changed the title [-]Updating to gemini 2.5 I get a GenerateContentResponse error[/-] [+]Updating to gemini 2.5 flash I get a GenerateContentResponse error[/+] on Jan 28, 2026
3. TimIgoe commented on Jul 20, 2026

   

   TimIgoe

   on Jul 20, 2026

   More actions

   Did anyone work anymore out on this, I've seen a massive increase in these responses over the weekend (same queries as were running last week - are now failing wiht that, but no useful data at all in teh response)
4. aydinfatih commented on Oct 12, 2026

   

   aydinfatih

   on Oct 12, 2026

   Member

   More actions

   This is usually caused by thinking: gemini-2.5-flash thinks before it answers, and the thinking tokens count towards maxOutputTokens. When the limit is low, the model can use it all on thoughts and stop with finishReason: MAX_TOKENS before writing an answer, so there are no parts to return.

   As of 2.8.0:

   • the text() / parts() error message includes the reason, e.g. ... Finish reason: MAX_TOKENS. (fix: Include finish and block reasons in quick accessor errors #189)
   • $response->candidates[0]->finishMessage is available when the API explains why the model stopped (feat: Add response metadata fields #179)

   To avoid it, increase maxOutputTokens or limit thinking:

   new GenerationConfig(
       maxOutputTokens: 2048,
       thinkingConfig: new ThinkingConfig(includeThoughts: false, thinkingBudget: 512),
   );

   See Empty Responses in the README.
5. aydinfatih

   closed this as completedon Oct 12, 2026