diff --git a/llama.cpp b/llama.cpp
index 8b675ea99..d9f30a26d 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -11957,7 +11957,7 @@ static size_t llama_output_reserve(llama_context & lctx, size_t n_outputs) {
     const auto n_embd  = hparams.n_embd;
 
     // TODO: use a per-batch flag for logits presence instead
-    const bool has_logits = cparams.causal_attn;
+    const bool has_logits = hparams.causal_attn;
     const bool has_embd   = cparams.embeddings && (hparams.causal_attn || cparams.pooling_type == LLAMA_POOLING_TYPE_NONE);
 
     const size_t logits_size = has_logits ? n_vocab*n_outputs_max : 0;