diff --git a/packages/ai/src/providers/openai-completions.ts b/packages/ai/src/providers/openai-completions.ts
index ebc2a2ff..8972e84e 100644
--- a/packages/ai/src/providers/openai-completions.ts
+++ b/packages/ai/src/providers/openai-completions.ts
@@ -369,15 +369,6 @@ function convertMessages(
 	let lastRole: string | null = null;
 
 	for (const msg of transformedMessages) {
-		// Some providers (e.g. Mistral) don't allow user messages directly after tool results
-		// Insert a synthetic assistant message to bridge the gap
-		if (compat.requiresAssistantAfterToolResult && lastRole === "toolResult" && msg.role === "user") {
-			params.push({
-				role: "assistant",
-				content: "I have processed the tool results.",
-			});
-		}
-
 		if (msg.role === "user") {
 			if (typeof msg.content === "string") {
 				params.push({
@@ -455,7 +446,16 @@ function convertMessages(
 					},
 				}));
 			}
-			if (assistantMsg.content === null && !assistantMsg.tool_calls) {
+			// Skip assistant messages that have no content and no tool calls.
+			// Mistral explicitly requires "either content or tool_calls, but not none".
+			// Other providers also don't accept empty assistant messages.
+			// This handles aborted assistant responses that got no content.
+			const content = assistantMsg.content;
+			const hasContent =
+				content !== null &&
+				content !== undefined &&
+				(typeof content === "string" ? content.length > 0 : content.length > 0);
+			if (!hasContent && !assistantMsg.tool_calls) {
 				continue;
 			}
 			params.push(assistantMsg);
@@ -570,7 +570,7 @@ function detectCompatFromUrl(baseUrl: string): Required<OpenAICompat> {
 		supportsReasoningEffort: !isGrok,
 		maxTokensField: useMaxTokens ? "max_tokens" : "max_completion_tokens",
 		requiresToolResultName: isMistral,
-		requiresAssistantAfterToolResult: isMistral,
+		requiresAssistantAfterToolResult: false, // Mistral no longer requires this as of Dec 2024
 		requiresThinkingAsText: isMistral,
 		requiresMistralToolIds: isMistral,
 	};
diff --git a/packages/ai/test/mistral-empty-assistant.test.ts b/packages/ai/test/mistral-empty-assistant.test.ts
new file mode 100644
index 00000000..fe037c0c
--- /dev/null
+++ b/packages/ai/test/mistral-empty-assistant.test.ts
@@ -0,0 +1,127 @@
+import { Mistral } from "@mistralai/mistralai";
+import { Type } from "@sinclair/typebox";
+import { describe, expect, it } from "vitest";
+import { getModel } from "../src/models.js";
+import { streamSimple } from "../src/stream.js";
+import type { AssistantMessage, Context, ToolCall, ToolResultMessage, UserMessage } from "../src/types.js";
+
+describe.skipIf(!process.env.MISTRAL_API_KEY)("Mistral Empty Assistant Message", () => {
+	it("verifies SDK rejects empty assistant messages", async () => {
+		// Verify the raw SDK behavior - empty assistant messages fail
+		const client = new Mistral({ apiKey: process.env.MISTRAL_API_KEY });
+
+		// This should fail - empty assistant message
+		try {
+			await client.chat.complete({
+				model: "devstral-medium-latest",
+				messages: [
+					{ role: "user", content: "Hello" },
+					{ role: "assistant", content: "" }, // Empty - should fail
+					{ role: "user", content: "Are you there?" },
+				],
+			});
+			expect.fail("Should have thrown an error");
+		} catch (error: any) {
+			expect(error.message).toContain("Assistant message must have either content or tool_calls");
+		}
+	});
+
+	it("skips empty assistant messages to avoid 400 errors", async () => {
+		const model = getModel("mistral", "devstral-medium-latest");
+		if (!model) throw new Error("Model not found");
+
+		// Build a context with an aborted assistant message
+		const messages: (UserMessage | AssistantMessage | ToolResultMessage)[] = [
+			{
+				role: "user",
+				content: "Hello, read a file for me",
+				timestamp: Date.now(),
+			},
+			{
+				role: "assistant",
+				content: [
+					{
+						type: "toolCall",
+						id: "test12345",
+						name: "read",
+						arguments: { path: "/test.txt" },
+					} as ToolCall,
+				],
+				api: "openai-completions",
+				provider: "mistral",
+				model: "devstral-medium-latest",
+				usage: {
+					input: 100,
+					output: 20,
+					cacheRead: 0,
+					cacheWrite: 0,
+					totalTokens: 120,
+					cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+				},
+				stopReason: "toolUse",
+				timestamp: Date.now(),
+			},
+			{
+				role: "toolResult",
+				toolCallId: "test12345",
+				toolName: "read",
+				content: [{ type: "text", text: "File content here..." }],
+				isError: false,
+				timestamp: Date.now(),
+			},
+			// This is the aborted assistant message - empty content, no tool calls
+			{
+				role: "assistant",
+				content: [], // Empty - simulates aborted
+				api: "openai-completions",
+				provider: "mistral",
+				model: "devstral-medium-latest",
+				usage: {
+					input: 0,
+					output: 0,
+					cacheRead: 0,
+					cacheWrite: 0,
+					totalTokens: 0,
+					cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+				},
+				stopReason: "aborted",
+				timestamp: Date.now(),
+				errorMessage: "Request was aborted.",
+			},
+			{
+				role: "user",
+				content: "Are you still there?",
+				timestamp: Date.now(),
+			},
+		];
+
+		const context: Context = {
+			systemPrompt: "You are a helpful assistant.",
+			messages,
+			tools: [
+				{
+					name: "read",
+					description: "Read file contents",
+					parameters: Type.Object({
+						path: Type.String(),
+					}),
+				},
+			],
+		};
+
+		// This should NOT fail with 400 after our fix
+		const response = await streamSimple(model, context);
+		const result = await response.result();
+
+		console.log("Result:", JSON.stringify(result, null, 2));
+
+		expect(result.stopReason).not.toBe("error");
+		expect(result.errorMessage).toBeUndefined();
+
+		// Verify the assistant can respond
+		const textContent = result.content.find((c) => c.type === "text");
+		expect(textContent).toBeDefined();
+
+		console.log("Test passed - pi-ai provider handled aborted message correctly");
+	}, 60000);
+});