diff --git a/src/api/providers/openai.ts b/src/api/providers/openai.ts
index 9d632fbdf4c..2949099bfe2 100644
--- a/src/api/providers/openai.ts
+++ b/src/api/providers/openai.ts
@@ -122,7 +122,10 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 					}
 				}
 
-				convertedMessages = [systemMessage, ...convertToOpenAiMessages(messages)]
+				// Use mergeToolResultText to merge text content into the last tool message
+				// This prevents "Unexpected role 'user' after role 'tool'" errors with
+				// strict OpenAI-compatible servers (vLLM, llama.cpp, Ollama with Mistral, etc.)
+				convertedMessages = [systemMessage, ...convertToOpenAiMessages(messages, { mergeToolResultText: true })]
 
 				if (modelInfo.supportsPromptCache) {
 					// Note: the following logic is copied from openrouter:
@@ -225,11 +228,13 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				yield this.processUsageMetrics(lastUsage, modelInfo)
 			}
 		} else {
+			// Use mergeToolResultText for non-streaming mode as well to prevent
+			// "Unexpected role 'user' after role 'tool'" errors with strict servers
 			const requestOptions: OpenAI.Chat.Completions.ChatCompletionCreateParamsNonStreaming = {
 				model: modelId,
 				messages: deepseekReasoner
 					? convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
-					: [systemMessage, ...convertToOpenAiMessages(messages)],
+					: [systemMessage, ...convertToOpenAiMessages(messages, { mergeToolResultText: true })],
 				...(metadata?.tools && { tools: this.convertToolsForOpenAI(metadata.tools) }),
 				...(metadata?.tool_choice && { tool_choice: metadata.tool_choice }),
 				...(metadata?.toolProtocol === "native" &&
@@ -344,6 +349,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 		if (this.options.openAiStreamingEnabled ?? true) {
 			const isGrokXAI = this._isGrokXAI(this.options.openAiBaseUrl)
 
+			// Use mergeToolResultText for O3 family models as well
 			const requestOptions: OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming = {
 				model: modelId,
 				messages: [
@@ -351,7 +357,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 						role: "developer",
 						content: `Formatting re-enabled\n${systemPrompt}`,
 					},
-					...convertToOpenAiMessages(messages),
+					...convertToOpenAiMessages(messages, { mergeToolResultText: true }),
 				],
 				stream: true,
 				...(isGrokXAI ? {} : { stream_options: { include_usage: true } }),
@@ -382,6 +388,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 
 			yield* this.handleStreamResponse(stream)
 		} else {
+			// Use mergeToolResultText for O3 family non-streaming mode as well
 			const requestOptions: OpenAI.Chat.Completions.ChatCompletionCreateParamsNonStreaming = {
 				model: modelId,
 				messages: [
@@ -389,7 +396,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 						role: "developer",
 						content: `Formatting re-enabled\n${systemPrompt}`,
 					},
-					...convertToOpenAiMessages(messages),
+					...convertToOpenAiMessages(messages, { mergeToolResultText: true }),
 				],
 				reasoning_effort: modelInfo.reasoningEffort as "low" | "medium" | "high" | undefined,
 				temperature: undefined,