diff --git a/src/models.js b/src/models.js
index 8190c0d..30f6dd9 100644
--- a/src/models.js
+++ b/src/models.js
@@ -116,6 +116,7 @@ export async function fetchModelsFromEndpoints() {
 					endpointId: endpoint.id,
 					uniqueId: `${endpoint.id}-${model.id}`,
 					maxTokens: maxTokens || 8192,
+					maxCompletionTokens: model.max_completion_tokens || model.max_output_tokens || maxTokens || 8192,
 				};
 			});
 		} catch (error) {
@@ -134,6 +135,7 @@ export async function fetchModelsFromEndpoints() {
 				url: endpoint.url,
 				endpointId: model.endpointId,
 				maxTokens: model.maxTokens,
+				maxCompletionTokens: model.maxCompletionTokens,
 				modelName: model.modelName,
 			};
 		}
@@ -325,6 +327,26 @@ export function getSelectedModelMaxTokens() {
 	return DEFAULT_MAX_TOKENS;
 }
 
+// Get max completion tokens for the currently selected model
+export function getSelectedModelMaxCompletionTokens() {
+	// Check modal dropdown first
+	const modalDropdown = document.getElementById("modal-model-selection");
+	if (modalDropdown?.value && modelRegistry[modalDropdown.value]) {
+		const modelData = modelRegistry[modalDropdown.value];
+		return modelData.maxCompletionTokens || DEFAULT_MAX_TOKENS;
+	}
+
+	// Fallback to main dropdown
+	const dropdown = document.getElementById("model-selection");
+	if (dropdown?.value && modelRegistry[dropdown.value]) {
+		const modelData = modelRegistry[dropdown.value];
+		return modelData.maxCompletionTokens || DEFAULT_MAX_TOKENS;
+	}
+
+	// Default fallback
+	return DEFAULT_MAX_TOKENS;
+}
+
 // Helper function to get the API endpoint for the selected model.
 export function getSelectedModelEndpoint() {
 	// Check modal dropdown first
diff --git a/src/translation.js b/src/translation.js
index ed1585c..e6b52a0 100644
--- a/src/translation.js
+++ b/src/translation.js
@@ -1,5 +1,5 @@
 import { API_KEY, getAPIConfig } from "./config.js";
-import { getSelectedModel, getSelectedModelEndpoint, getSelectedModelMaxTokens, modelRegistry } from "./models.js";
+import { getSelectedModel, getSelectedModelEndpoint, getSelectedModelMaxTokens, getSelectedModelMaxCompletionTokens, modelRegistry } from "./models.js";
 import { countTokens, countTokensWithUpstream } from "./token_estimator.js";
 
 // Languages supported by Hermes 3 model (in English for reference)
@@ -77,7 +77,12 @@ async function* sendMessageWithHistory(messageHistory) {
 	const inputText = messageHistory.map(msg => msg.content).join('');
 	const inputTokens = Math.ceil(inputText.length / 4); // Simple token estimation
 	const outputReserve = Math.floor(modelMaxTokens * 0.4); // Reserve 40% for output
-	const availableTokens = Math.max(512, modelMaxTokens - inputTokens - outputReserve);
+	let availableTokens = Math.max(512, modelMaxTokens - inputTokens - outputReserve);
+	
+	// Cap output tokens based on model's max completion tokens limit
+	const maxCompletionTokens = getSelectedModelMaxCompletionTokens();
+	availableTokens = Math.min(availableTokens, maxCompletionTokens);
+	console.log(`Capped available tokens to model's max completion limit: ${maxCompletionTokens}`);
 	
 	console.log("Translation using endpoint:", apiUrl);
 	console.log("Translation using model:", model);