diff --git a/gpt4all-backend/falcon.cpp b/gpt4all-backend/falcon.cpp
index cdb135a1..8fe99c3a 100644
--- a/gpt4all-backend/falcon.cpp
+++ b/gpt4all-backend/falcon.cpp
@@ -430,7 +430,7 @@ bool falcon_model_load(const std::string & fname, falcon_model & model, gpt_voca
 
     fin.close();
 
-    model.eval_buf.resize(256u * 1024 * 1024);
+    model.eval_buf.resize(1280u * 1024 * 1024);
     model.scr0_buf.resize(256u * 1024 * 1024);
     model.scr1_buf.resize(256u * 1024 * 1024);
     return true;
diff --git a/gpt4all-backend/replit.cpp b/gpt4all-backend/replit.cpp
index 821100a5..6f7c0a52 100644
--- a/gpt4all-backend/replit.cpp
+++ b/gpt4all-backend/replit.cpp
@@ -485,7 +485,7 @@ bool replit_model_load(const std::string & fname, std::istream &fin, replit_mode
         printf("%s: model size = %8.2f MB / num tensors = %d\n", __func__, total_size / 1024.0 / 1024.0, n_tensors);
     }
 
-   model.eval_buf.resize(256u * 1024 * 1024);
+   model.eval_buf.resize(512u * 1024 * 1024);
    model.scr0_buf.resize(256u * 1024 * 1024);
    model.scr1_buf.resize(256u * 1024 * 1024);