diff --git a/src/chat.js b/src/chat.js
index 353e747..400144c 100644
--- a/src/chat.js
+++ b/src/chat.js
@@ -110,15 +110,10 @@ export async function* sendMessage(message) {
 	
 	// Cap to model's max completion tokens limit
 	const maxCompletionTokens = getSelectedModelMaxCompletionTokens();
-	// For chat, prefer using the full calculated available space for coherent responses
-	// Only cap if we're exceeding the model's hard limit
+	// Use the full calculated available space or the model's limit, whichever is smaller
 	const finalAvailable = Math.min(calculatedAvailable, maxCompletionTokens);
 	
-	// However, for very large available token counts, cap at a reasonable max to prevent rambling
-	const reasonableMax = 4096; // Enough for very thorough responses without being excessive
-	const actualFinal = Math.min(finalAvailable, reasonableMax);
-	
-	console.log("Chat using tokens:", actualFinal, "calculated:", calculatedAvailable, "max completion:", maxCompletionTokens);
+	console.log("Chat using tokens:", finalAvailable, "calculated:", calculatedAvailable, "max completion:", maxCompletionTokens);
 	
 	const response = await fetch(apiUrl, {
 		method: "POST",
@@ -127,7 +122,7 @@ export async function* sendMessage(message) {
 			model: model,
 			messages: chatHistory,
 			temperature: 0.5,
-			max_tokens: actualFinal,
+			max_tokens: finalAvailable,
 			stream: true,
 		}),
 	});