diff --git a/src/modules/aix/server/dispatch/chatGenerate/parsers/openai.parser.ts b/src/modules/aix/server/dispatch/chatGenerate/parsers/openai.parser.ts
index 323f52045..163177b43 100644
--- a/src/modules/aix/server/dispatch/chatGenerate/parsers/openai.parser.ts
+++ b/src/modules/aix/server/dispatch/chatGenerate/parsers/openai.parser.ts
@@ -410,7 +410,7 @@ function _fromOpenAIUsage(usage: OpenAIWire_API_Chat_Completions.Response['usage
   // Input Metrics
 
   // Input redistribution: Cache Read
-  if (usage.prompt_tokens_details !== undefined) {
+  if (usage.prompt_tokens_details) {
     const TCacheRead = usage.prompt_tokens_details.cached_tokens;
     if (TCacheRead !== undefined && TCacheRead > 0) {
       metricsUpdate.TCacheRead = TCacheRead;
@@ -434,8 +434,11 @@ function _fromOpenAIUsage(usage: OpenAIWire_API_Chat_Completions.Response['usage
   // Output Metrics
 
   // Output breakdown: Reasoning
-  if (usage.completion_tokens_details?.reasoning_tokens !== undefined)
-    metricsUpdate.TOutR = usage.completion_tokens_details.reasoning_tokens;
+  if (usage.completion_tokens_details) {
+    const details = usage.completion_tokens_details || {};
+    if (details.reasoning_tokens !== undefined)
+      metricsUpdate.TOutR = usage.completion_tokens_details.reasoning_tokens;
+  }
 
   // TODO: Output breakdown: Audio
 
diff --git a/src/modules/aix/server/dispatch/wiretypes/openai.wiretypes.ts b/src/modules/aix/server/dispatch/wiretypes/openai.wiretypes.ts
index 99c25a0e8..8e87d7eed 100644
--- a/src/modules/aix/server/dispatch/wiretypes/openai.wiretypes.ts
+++ b/src/modules/aix/server/dispatch/wiretypes/openai.wiretypes.ts
@@ -394,7 +394,8 @@ export namespace OpenAIWire_API_Chat_Completions {
     prompt_tokens_details: z.object({
       audio_tokens: z.number().optional(),
       cached_tokens: z.number().optional(),
-    }).optional(),
+    }).optional()
+      .nullable(), // [2025-06-02] Chutes.ai using slang server returns null for prompt_tokens_details
 
     // [OpenAI o1, 2024-09-12] breaks down the completion tokens into components
     completion_tokens_details: z.object({
@@ -403,7 +404,8 @@ export namespace OpenAIWire_API_Chat_Completions {
       audio_tokens: z.number().optional(), // [OpenAI, 2024-10-01] audio tokens used in the completion (charged at a different rate)
       accepted_prediction_tokens: z.number().optional(), // [OpenAI, 2024-11-05] Predicted Outputs
       rejected_prediction_tokens: z.number().optional(), // [OpenAI, 2024-11-05] Predicted Outputs
-    }).optional(), // not present in other APIs yet
+    }).optional() // not present in other APIs yet
+      .nullable(), // [2025-06-02] no issues yet, but preventive
 
     // [DeepSeek, 2024-08-02] context caching on disk
     prompt_cache_hit_tokens: z.number().optional(),