diff --git a/changes/next_release.md b/changes/next_release.md
new file mode 100644
index 00000000..b0fd44bf
--- /dev/null
+++ b/changes/next_release.md
@@ -0,0 +1,4 @@
+
+**Recommendation:** Because `GPT-4o-mini` is much smarter than `GPT-3.5-turbo`, I recommend you to disable GPT function processor `FFUN_LIBRARIAN_OPENAI_CHAT_35_FUNCTIONS_PROCESSOR__ENABLED="False"` and use only standard GPT processor `FFUN_LIBRARIAN_OPENAI_CHAT_35_PROCESSOR__ENABLED="False"`. In [gh-227](https://github.com/Tiendil/feeds.fun/issues/227), one of the processors will be refactored, and one of them will be removed.
+
+- [gh-226](https://github.com/Tiendil/feeds.fun/issues/226) Switch used OpenAI model to gpt4o-mini
diff --git a/changes/unreleased.md b/changes/unreleased.md
deleted file mode 100644
index 373eb997..00000000
--- a/changes/unreleased.md
+++ /dev/null
@@ -1,2 +0,0 @@
-
-No changes.
diff --git a/ffun/ffun/application/user_settings.py b/ffun/ffun/application/user_settings.py
index 3c7571c5..093eeb30 100644
--- a/ffun/ffun/application/user_settings.py
+++ b/ffun/ffun/application/user_settings.py
@@ -32,8 +32,8 @@ class UserSetting(enum.IntEnum):
 The more users set up the key, the cheaper it will be for everyone.
 """
 
-openai_max_token_cost = decimal.Decimal("0.004")
-openai_max_token_cost_n = 1000
+openai_max_token_cost = (decimal.Decimal("0.150") + decimal.Decimal("0.600")) / 2
+openai_max_token_cost_n = 1000000
 openai_max_spendings = decimal.Decimal("10.00")
 
 description_openai_max_tokens_in_month = f"""
diff --git a/ffun/ffun/librarian/processors/openai_chat_3_5.py b/ffun/ffun/librarian/processors/openai_chat_3_5.py
index a4723c34..767494f8 100644
--- a/ffun/ffun/librarian/processors/openai_chat_3_5.py
+++ b/ffun/ffun/librarian/processors/openai_chat_3_5.py
@@ -17,10 +17,10 @@
 
 system = """\
 You are an expert on the analysis of text semantics.
-For provided text, you determine a list of best tags to describe the text.
-For each category, you provide 30 tags.
+For the provided text, you determine a list of best tags to describe the text.
+For each category, you provide up to 30 tags.
 
-Categories are topics, meta-topics, high-level-topics, low-level-topics, related-topics,\
+Categories are topics, meta-topics, high-level-topics, low-level-topics, related-topics, \
 indirect-topics, mentions, indirect-mentions.
 
 Tags are only in English. Normalize tags and output them as JSON.\
diff --git a/ffun/ffun/librarian/settings.py b/ffun/ffun/librarian/settings.py
index 7dd31989..407fee33 100644
--- a/ffun/ffun/librarian/settings.py
+++ b/ffun/ffun/librarian/settings.py
@@ -21,12 +21,14 @@ class UpperCaseTitleProcessor(BaseProcessor):
     pass
 
 
+# TODO: will be ranamed & refactored in gh-227
 class OpenAIChat35Processor(BaseProcessor):
-    model: str = "gpt-3.5-turbo-0125"
+    model: str = "gpt-4o-mini-2024-07-18"
 
 
+# TODO: will be removed in gh-227
 class OpenAIChat35FunctionsProcessor(BaseProcessor):
-    model: str = "gpt-3.5-turbo-0125"
+    model: str = "gpt-4o-mini-2024-07-18"
 
 
 class Settings(BaseSettings):