diff --git a/pkg/ollama/http_handler.go b/pkg/ollama/http_handler.go
index 2d4b1620..35372399 100644
--- a/pkg/ollama/http_handler.go
+++ b/pkg/ollama/http_handler.go
@@ -373,13 +373,6 @@ func (h *HTTPHandler) handleChat(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	// Handle num_ctx option for context size configuration
-	if req.Options != nil {
-		if numCtxRaw, ok := req.Options["num_ctx"]; ok {
-			h.configure(r.Context(), numCtxRaw, sanitizedModelName, modelName, r.UserAgent())
-		}
-	}
-
 	// Convert to OpenAI format chat completion request
 	openAIReq := map[string]interface{}{
 		"model":    modelName,
@@ -387,14 +380,12 @@ func (h *HTTPHandler) handleChat(w http.ResponseWriter, r *http.Request) {
 		"stream":   req.Stream == nil || *req.Stream,
 	}
 
-	// Add options if present
 	if req.Options != nil {
-		if temp, ok := req.Options["temperature"]; ok {
-			openAIReq["temperature"] = temp
-		}
-		if maxTokens, ok := req.Options["num_predict"]; ok {
-			openAIReq["max_tokens"] = maxTokens
+		// Handle num_ctx option for context size configuration
+		if numCtxRaw, ok := req.Options["num_ctx"]; ok {
+			h.configure(r.Context(), numCtxRaw, sanitizedModelName, modelName, r.UserAgent())
 		}
+		h.mapOllamaOptionsToOpenAI(req.Options, openAIReq)
 	}
 
 	// Make request to scheduler
@@ -465,14 +456,9 @@ func (h *HTTPHandler) handleGenerate(w http.ResponseWriter, r *http.Request) {
 		"stream": req.Stream == nil || *req.Stream,
 	}
 
-	// Add options if present
+	// Map Ollama options to OpenAI format
 	if req.Options != nil {
-		if temp, ok := req.Options["temperature"]; ok {
-			openAIReq["temperature"] = temp
-		}
-		if maxTokens, ok := req.Options["num_predict"]; ok {
-			openAIReq["max_tokens"] = maxTokens
-		}
+		h.mapOllamaOptionsToOpenAI(req.Options, openAIReq)
 	}
 
 	// Make request to scheduler
@@ -671,6 +657,39 @@ func (h *HTTPHandler) handlePull(w http.ResponseWriter, r *http.Request) {
 	}
 }
 
+// mapOllamaOptionsToOpenAI maps Ollama API options to OpenAI-compatible format
+// This function handles all standard Ollama options and maps them to their OpenAI equivalents
+func (h *HTTPHandler) mapOllamaOptionsToOpenAI(ollamaOpts map[string]interface{}, openAIReq map[string]interface{}) {
+	// Direct mappings - these options have exact OpenAI equivalents
+	if val, ok := ollamaOpts["temperature"]; ok {
+		openAIReq["temperature"] = val
+	}
+	if val, ok := ollamaOpts["top_p"]; ok {
+		openAIReq["top_p"] = val
+	}
+	if val, ok := ollamaOpts["top_k"]; ok {
+		openAIReq["top_k"] = val
+	}
+	if val, ok := ollamaOpts["num_predict"]; ok {
+		openAIReq["max_tokens"] = val
+	}
+	if val, ok := ollamaOpts["stop"]; ok {
+		openAIReq["stop"] = val
+	}
+	if val, ok := ollamaOpts["seed"]; ok {
+		openAIReq["seed"] = val
+	}
+	if val, ok := ollamaOpts["presence_penalty"]; ok {
+		openAIReq["presence_penalty"] = val
+	}
+	if val, ok := ollamaOpts["frequency_penalty"]; ok {
+		openAIReq["frequency_penalty"] = val
+	}
+
+	// Note: num_ctx is handled separately in the configure() function
+	// as it requires a special ConfigureRunner call
+}
+
 // convertMessages converts Ollama messages to OpenAI format
 func convertMessages(messages []Message) []map[string]interface{} {
 	result := make([]map[string]interface{}, len(messages))