Google Gemini HTTP 400 Bad request

Gemini 400 — Function call missing a thought_signature

What this error means

Gemini 2.5 returns 400 INVALID_ARGUMENT when a functionCall part is replayed without the thought_signature it originally issued.

How to fix it

Preserve thought_signature on every functionCall part you send back
Keep function turns immediately after the matching user or tool turn
Do not hand-edit or truncate prior tool messages

Example error message

{
  "error": {
    "code": 400,
    "message": "Function call is missing a thought_signature in functionCall parts.",
    "status": "INVALID_ARGUMENT"
  }
}

Frequently asked

Why did this start after upgrading to Gemini 2.5?

Thought signatures are new to 2.5 thinking models; older clients that strip unknown fields drop them and trigger the 400.

More Google Gemini errors

Code	Error	Affected teams / 30 days
429	Gemini 429 — You exceeded your current quota	120	View
503	Gemini 503 — The model is currently overloaded	40	View
View all Google Gemini errors

Route Google Gemini through Manifest Live token usage, every model, and automatic fallback when Google Gemini rate-limits or fails.

MANIFEST AUTO-FIX

Don't let your requests fail again and again

Manifest fixes your bad LLM requests on the fly so they return successful responses before it reaches your agent. No downtime.

Deprecated / Not-found models
Wrong parameters
Malformed requests
Exceeded context windows

Get started Talk to sales

All errors

Provider	Code	Error	Type	Affected teams / 30 days
Google Gemini	429	Gemini 429 — You exceeded your current quota	Rate limit	120	View
OpenRouter	404	OpenRouter 404 — No endpoints found for model	Model unavailable	80	View
OpenRouter	402	OpenRouter 402 — Insufficient credits	Billing / quota	70	View
OpenAI	429	OpenAI 429 — Rate limit exceeded	Rate limit	68	View
OpenAI	401	OpenAI 401 — Invalid authentication	Authentication	55	View
DeepSeek	400	DeepSeek 400 — reasoning_content must be passed	Bad request	48	View
Google Gemini	400	Gemini 400 — Function call missing a thought_signature	Bad request	42	View
Google Gemini	503	Gemini 503 — The model is currently overloaded	Server error	40	View
Groq	429	Groq 429 — Rate limit exceeded	Rate limit	38	View
OpenAI	400	OpenAI 400 — Context length exceeded	Bad request	34	View
Anthropic	400	Anthropic 400 — You are out of extra usage	Bad request	30	View
Anthropic	529	Anthropic 529 — Overloaded	Server error	26	View
Browse all errors