spring-projects/spring-ai · error · IllegalArgumentException

ThinkingLevel.%s is not supported for model '%s'. Supported

Error message

ThinkingLevel.%s is not supported for model '%s'. Supported levels: %s.

What it means

Model-capability guard in GoogleGenAiChatModel.validateThinkingLevelForModel: the requested GoogleGenAiThinkingLevel is not in the per-model supported set looked up by model id (Vertex-style resource names are reduced to the last path segment), so the request would fail or be ignored by that model.

Source

Thrown at models/spring-ai-google-genai/src/main/java/org/springframework/ai/google/genai/GoogleGenAiChatModel.java:930

	private static void validateThinkingLevelForModel(GoogleGenAiThinkingLevel level, String modelName) {
		if (level == null || level == GoogleGenAiThinkingLevel.THINKING_LEVEL_UNSPECIFIED || modelName == null) {
			return;
		}
		// Vertex AI style full resource names (e.g.
		// "projects/{project}/locations/{location}/publishers/google/models/{model}")
		// carry the model id as the last path segment.
		String modelId = modelName.substring(modelName.lastIndexOf('/') + 1).toLowerCase(Locale.ROOT);
		Set<GoogleGenAiThinkingLevel> supportedLevels = THINKING_LEVEL_SUPPORT_BY_MODEL.get(modelId);
		if (supportedLevels == null) {
			return;
		}
		if (!supportedLevels.contains(level)) {
			if (supportedLevels.isEmpty()) {
				throw new IllegalArgumentException(String.format(
						"ThinkingLevel.%s is not supported for model '%s'. This model does not support thinkingLevel; use thinkingBudget instead.",
						level, modelName));
			}
			throw new IllegalArgumentException(
					String.format("ThinkingLevel.%s is not supported for model '%s'. Supported levels: %s.", level,
							modelName, supportedLevels.stream().map(Enum::name).collect(Collectors.joining(", "))));
		}
	}

	private List<Content> toGeminiContent(List<Message> instructions) {

		List<Content> contents = instructions.stream()
			.map(message -> Content.builder()
				.role(toGeminiMessageType(message.getMessageType()).getValue())
				.parts(messageToGeminiParts(message))
				.build())
			.toList();

		return contents;
	}

	private List<SafetySetting> toGeminiSafetySettings(List<GoogleGenAiSafetySetting> safetySettings) {

View on GitHub (pinned to 98a7beda4f)

Solutions

Use a model that supports the requested thinking level
Omit thinkingLevel or use THINKING_LEVEL_UNSPECIFIED
Pick a level from the supported-levels list in the message

Example fix

// before
options.thinkingLevel(GoogleGenAiThinkingLevel.THINKING_LEVEL_HIGH) // on a LOW/MEDIUM-only model
// after
options.thinkingLevel(GoogleGenAiThinkingLevel.THINKING_LEVEL_MEDIUM)

Defensive patterns

Strategy: validation

Validate before calling

// validate against the level list reported by the error/docs
if (!Set.of(THINKING_LEVEL_LOW, THINKING_LEVEL_MEDIUM).contains(level)) {
    throw new IllegalArgumentException("level not supported for this model");
}

Try / catch

try { model.call(prompt); } catch (IllegalArgumentException e) { if (e.getMessage().contains("Supported levels:")) pickSupportedLevel(e.getMessage()); else throw e; }

Prevention

Read the Supported levels list in the message and adjust
Don't share one options template across all Gemini variants
Pin model IDs and tune thinking levels per model

When it happens

Trigger: Requesting a ThinkingLevel (e.g. THINKING_LEVEL_HIGH) on a model whose THINKING_LEVEL_SUPPORT_BY_MODEL entry contains other levels only (e.g. a model supporting only THINKING_LEVEL_LOW and THINKING_LEVEL_MEDIUM).

Common situations: Sharing options templates across Gemini 2.5 variants with differing level support; hardcoding a level tuned for Pro on Flash models; after changing model IDs in configuration.

Related errors

AI-assisted analysis of spring-projects/spring-ai@98a7beda4f (2026-09-11). Data as JSON: /api/errors/5df58b7534c55936. Report an issue: GitHub.