diff --git a/backend/.env.example b/backend/.env.example
index 6d74639..8fa7162 100644
--- a/backend/.env.example
+++ b/backend/.env.example
@@ -14,5 +14,5 @@ PORT=8080
 ALLOWED_ORIGINS=http://localhost:3000
 
 # External APIs
-GEMINI_API_KEY=your-gemini-key
+OPENAI_API_KEY=your-openai-key
 PEXELS_API_KEY=your-pexels-key
diff --git a/backend/cmd/server/main.go b/backend/cmd/server/main.go
index 032bd16..ca9c297 100644
--- a/backend/cmd/server/main.go
+++ b/backend/cmd/server/main.go
@@ -93,7 +93,7 @@ func run() error {
 	authMW := middleware.Auth(&jwtAdapter{jm: jwtManager})
 
 	// External API clients
-	geminiClient := gemini.NewClient(cfg.GeminiAPIKey)
+	geminiClient := gemini.NewClient(cfg.OpenAIAPIKey)
 	pexelsClient := pexels.NewClient(cfg.PexelsAPIKey)
 
 	// Ingredient domain
diff --git a/backend/internal/config/config.go b/backend/internal/config/config.go
index 2b5f530..b7ef54b 100644
--- a/backend/internal/config/config.go
+++ b/backend/internal/config/config.go
@@ -22,7 +22,7 @@ type Config struct {
 	AllowedOrigins []string `envconfig:"ALLOWED_ORIGINS" default:"http://localhost:3000"`
 
 	// External APIs
-	GeminiAPIKey  string `envconfig:"GEMINI_API_KEY" required:"true"`
+	OpenAIAPIKey  string `envconfig:"OPENAI_API_KEY" required:"true"`
 	PexelsAPIKey  string `envconfig:"PEXELS_API_KEY" required:"true"`
 }
 
diff --git a/backend/internal/gemini/client.go b/backend/internal/gemini/client.go
index 96788cf..f89d5db 100644
--- a/backend/internal/gemini/client.go
+++ b/backend/internal/gemini/client.go
@@ -11,19 +11,19 @@ import (
 )
 
 const (
-	// groqAPIURL is the Groq OpenAI-compatible endpoint (free tier, no billing required).
-	groqAPIURL = "https://api.groq.com/openai/v1/chat/completions"
+	// openaiAPIURL is the OpenAI chat completions endpoint.
+	openaiAPIURL = "https://api.openai.com/v1/chat/completions"
 
-	// groqModel is the default text generation model.
-	groqModel = "llama-3.3-70b-versatile"
+	// openaiModel is the default text generation model.
+	openaiModel = "gpt-4o-mini"
 
-	// groqVisionModel supports image inputs in OpenAI vision format.
-	groqVisionModel = "meta-llama/llama-4-scout-17b-16e-instruct"
+	// openaiVisionModel supports image inputs.
+	openaiVisionModel = "gpt-4o"
 
 	maxRetries = 3
 )
 
-// Client is an HTTP client for the Groq LLM API (OpenAI-compatible).
+// Client is an HTTP client for the OpenAI API.
 type Client struct {
 	apiKey     string
 	httpClient *http.Client
@@ -39,9 +39,9 @@ func NewClient(apiKey string) *Client {
 	}
 }
 
-// generateContent sends text messages to the text-only model.
+// generateContent sends text messages to the text model.
 func (c *Client) generateContent(ctx context.Context, messages []map[string]string) (string, error) {
-	return c.callGroq(ctx, groqModel, 0.7, messages)
+	return c.callOpenAI(ctx, openaiModel, 0.7, messages)
 }
 
 // generateVisionContent sends an image + text prompt to the vision model.
@@ -68,12 +68,12 @@ func (c *Client) generateVisionContent(ctx context.Context, prompt, imageBase64,
 			},
 		},
 	}
-	return c.callGroq(ctx, groqVisionModel, 0.1, messages)
+	return c.callOpenAI(ctx, openaiVisionModel, 0.1, messages)
 }
 
-// callGroq is the shared HTTP transport for all Groq requests.
+// callOpenAI is the shared HTTP transport for all OpenAI requests.
 // messages can be []map[string]string (text) or []any (vision with image content).
-func (c *Client) callGroq(ctx context.Context, model string, temperature float64, messages any) (string, error) {
+func (c *Client) callOpenAI(ctx context.Context, model string, temperature float64, messages any) (string, error) {
 	body := map[string]any{
 		"model":       model,
 		"temperature": temperature,
@@ -85,7 +85,7 @@ func (c *Client) callGroq(ctx context.Context, model string, temperature float64
 		return "", fmt.Errorf("marshal request: %w", err)
 	}
 
-	req, err := http.NewRequestWithContext(ctx, http.MethodPost, groqAPIURL, bytes.NewReader(bodyBytes))
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, openaiAPIURL, bytes.NewReader(bodyBytes))
 	if err != nil {
 		return "", fmt.Errorf("create request: %w", err)
 	}
@@ -100,7 +100,7 @@ func (c *Client) callGroq(ctx context.Context, model string, temperature float64
 
 	if resp.StatusCode != http.StatusOK {
 		raw, _ := io.ReadAll(resp.Body)
-		return "", fmt.Errorf("groq API error %d: %s", resp.StatusCode, string(raw))
+		return "", fmt.Errorf("openai API error %d: %s", resp.StatusCode, string(raw))
 	}
 
 	var result struct {
@@ -114,7 +114,7 @@ func (c *Client) callGroq(ctx context.Context, model string, temperature float64
 		return "", fmt.Errorf("decode response: %w", err)
 	}
 	if len(result.Choices) == 0 {
-		return "", fmt.Errorf("empty response from Groq")
+		return "", fmt.Errorf("empty response from OpenAI")
 	}
 	return result.Choices[0].Message.Content, nil
 }
diff --git a/backend/internal/gemini/recipe.go b/backend/internal/gemini/recipe.go
index ac642e8..ef6be73 100644
--- a/backend/internal/gemini/recipe.go
+++ b/backend/internal/gemini/recipe.go
@@ -68,7 +68,7 @@ type NutritionInfo struct {
 func (c *Client) GenerateRecipes(ctx context.Context, req RecipeRequest) ([]Recipe, error) {
 	prompt := buildRecipePrompt(req)
 
-	// OpenAI-compatible messages format used by Groq.
+	// OpenAI messages format.
 	messages := []map[string]string{
 		{"role": "user", "content": prompt},
 	}
diff --git a/backend/internal/menu/handler.go b/backend/internal/menu/handler.go
index 3a9416a..1c9bea0 100644
--- a/backend/internal/menu/handler.go
+++ b/backend/internal/menu/handler.go
@@ -135,7 +135,7 @@ func (h *Handler) GenerateMenu(w http.ResponseWriter, r *http.Request) {
 		menuReq.AvailableProducts = products
 	}
 
-	// Generate 7-day plan via Groq.
+	// Generate 7-day plan via OpenAI.
 	days, err := h.gemini.GenerateMenu(r.Context(), menuReq)
 	if err != nil {
 		slog.Error("generate menu", "user_id", userID, "err", err)