diff --git a/pkg/monitor/display.go b/pkg/monitor/display.go
index c84e9cdb..041c366f 100644
--- a/pkg/monitor/display.go
+++ b/pkg/monitor/display.go
@@ -288,7 +288,9 @@ func (d *display) Stop(output string, err error) {
 	defer d.callLock.Unlock()
 
 	log.Fields("runID", d.dump.ID, "output", output, "err", err).Debugf("Run stopped")
-	log.Fields("runID", d.dump.ID, "total", d.usage.TotalTokens, "prompt", d.usage.PromptTokens, "completion", d.usage.CompletionTokens).Infof("usage   ")
+	if d.usage.TotalTokens > 0 {
+		log.Fields("runID", d.dump.ID, "total", d.usage.TotalTokens, "prompt", d.usage.PromptTokens, "completion", d.usage.CompletionTokens).Infof("usage   ")
+	}
 	d.dump.Output = output
 	d.dump.Err = err
 	if d.dumpState != "" {
diff --git a/pkg/openai/client.go b/pkg/openai/client.go
index a04ce2d4..2c72bfa8 100644
--- a/pkg/openai/client.go
+++ b/pkg/openai/client.go
@@ -308,9 +308,6 @@ func (c *Client) Call(ctx context.Context, messageRequest types.CompletionReques
 		Model:     messageRequest.Model,
 		Messages:  msgs,
 		MaxTokens: messageRequest.MaxTokens,
-		StreamOptions: &openai.StreamOptions{
-			IncludeUsage: true,
-		},
 	}
 
 	if messageRequest.Temperature == nil {
@@ -350,6 +347,9 @@ func (c *Client) Call(ctx context.Context, messageRequest types.CompletionReques
 	var cacheResponse bool
 	if c.setSeed {
 		request.Seed = ptr(c.seed(request))
+		request.StreamOptions = &openai.StreamOptions{
+			IncludeUsage: true,
+		}
 	}
 	response, ok, err := c.fromCache(ctx, messageRequest, request)
 	if err != nil {