diff --git a/backend/open_webui/config.py b/backend/open_webui/config.py
index f89738b0db..19da0f561e 100644
--- a/backend/open_webui/config.py
+++ b/backend/open_webui/config.py
@@ -2128,6 +2128,10 @@ ENABLE_CONTEXT_COMPACTION = os.getenv('ENABLE_CONTEXT_COMPACTION', 'False').lowe
 
 CONTEXT_COMPACTION_TOKEN_THRESHOLD = int(os.getenv('CONTEXT_COMPACTION_TOKEN_THRESHOLD', '80000'))
 
+CONTEXT_COMPACTION_TOKEN_CAP = int(
+    os.getenv('CONTEXT_COMPACTION_TOKEN_CAP', os.getenv('CONTEXT_COMPACTION_TOKEN_THRESHOLD', '80000'))
+)
+
 CONTEXT_COMPACTION_PROMPT_TEMPLATE = os.getenv('CONTEXT_COMPACTION_PROMPT_TEMPLATE', '')
 
 TITLE_GENERATION_PROMPT_TEMPLATE = os.getenv('TITLE_GENERATION_PROMPT_TEMPLATE', '')
@@ -3034,6 +3038,7 @@ DEFAULT_CONFIG = {
     'task.model.external': TASK_MODEL_EXTERNAL,
     'chat.context_compaction.enable': ENABLE_CONTEXT_COMPACTION,
     'chat.context_compaction.token_threshold': CONTEXT_COMPACTION_TOKEN_THRESHOLD,
+    'chat.context_compaction.token_cap': CONTEXT_COMPACTION_TOKEN_CAP,
     'chat.context_compaction.prompt_template': CONTEXT_COMPACTION_PROMPT_TEMPLATE,
     'task.title.prompt_template': TITLE_GENERATION_PROMPT_TEMPLATE,
     'task.tags.prompt_template': TAGS_GENERATION_PROMPT_TEMPLATE,
diff --git a/backend/open_webui/routers/chats.py b/backend/open_webui/routers/chats.py
index 98621ff80e..d9f6e6706b 100644
--- a/backend/open_webui/routers/chats.py
+++ b/backend/open_webui/routers/chats.py
@@ -51,6 +51,7 @@ SEARCH_FILTER_PREFIXES = ('tag:', 'folder:', 'pinned:', 'archived:', 'shared:')
 CHAT_CONFIG_KEYS = {
     'ENABLE_CONTEXT_COMPACTION': 'chat.context_compaction.enable',
     'CONTEXT_COMPACTION_TOKEN_THRESHOLD': 'chat.context_compaction.token_threshold',
+    'CONTEXT_COMPACTION_TOKEN_CAP': 'chat.context_compaction.token_cap',
     'CONTEXT_COMPACTION_PROMPT_TEMPLATE': 'chat.context_compaction.prompt_template',
 }
 
@@ -58,6 +59,7 @@ CHAT_CONFIG_KEYS = {
 class ChatConfigForm(BaseModel):
     ENABLE_CONTEXT_COMPACTION: bool
     CONTEXT_COMPACTION_TOKEN_THRESHOLD: int
+    CONTEXT_COMPACTION_TOKEN_CAP: int
     CONTEXT_COMPACTION_PROMPT_TEMPLATE: str
 
 
@@ -712,11 +714,13 @@ async def get_chat_config(user=Depends(get_admin_user)):
 @router.post('/config', response_model=ChatConfigForm)
 async def set_chat_config(form_data: ChatConfigForm, user=Depends(get_admin_user)):
     threshold = max(1, int(form_data.CONTEXT_COMPACTION_TOKEN_THRESHOLD))
+    token_cap = max(1, int(form_data.CONTEXT_COMPACTION_TOKEN_CAP))
     await Config.upsert(
         chat_config_updates(
             {
                 **form_data.model_dump(),
                 'CONTEXT_COMPACTION_TOKEN_THRESHOLD': threshold,
+                'CONTEXT_COMPACTION_TOKEN_CAP': token_cap,
             }
         )
     )
diff --git a/backend/open_webui/utils/context_compaction.py b/backend/open_webui/utils/context_compaction.py
index 209f56892f..2cdc4c3aef 100644
--- a/backend/open_webui/utils/context_compaction.py
+++ b/backend/open_webui/utils/context_compaction.py
@@ -53,7 +53,7 @@ async def compact_messages_for_request(
         return messages, None, False
 
     messages, previous_summary = _apply_latest_summary_checkpoint(messages)
-    token_threshold = _resolve_token_threshold(config['token_threshold'], metadata)
+    token_threshold = _resolve_token_threshold(config['token_threshold'], config['token_cap'], metadata)
     if not _exceeds_token_threshold(messages, system_prompt, previous_summary, token_threshold) or len(messages) <= 3:
         return messages, previous_summary, False
 
@@ -186,11 +186,14 @@ async def _load_config() -> dict:
     values = await Config.get_many(
         'chat.context_compaction.enable',
         'chat.context_compaction.token_threshold',
+        'chat.context_compaction.token_cap',
         'chat.context_compaction.prompt_template',
     )
+    token_threshold = _parse_positive_int(values.get('chat.context_compaction.token_threshold')) or 80000
     return {
         'enable': bool(values.get('chat.context_compaction.enable', False)),
-        'token_threshold': int(values.get('chat.context_compaction.token_threshold', 80000) or 80000),
+        'token_threshold': token_threshold,
+        'token_cap': _parse_positive_int(values.get('chat.context_compaction.token_cap')) or token_threshold,
         'prompt_template': values.get('chat.context_compaction.prompt_template', '') or '',
     }
 
@@ -203,9 +206,9 @@ def _parse_positive_int(value: Any) -> int | None:
     return parsed if parsed > 0 else None
 
 
-def _resolve_token_threshold(global_threshold: int, metadata: dict) -> int:
+def _resolve_token_threshold(global_threshold: int, global_cap: int, metadata: dict) -> int:
     configured_threshold = _parse_positive_int((metadata.get('params') or {}).get('compact_token_threshold'))
-    return configured_threshold or global_threshold
+    return min(configured_threshold or global_threshold, global_cap)
 
 
 def _apply_latest_summary_checkpoint(messages: list[dict]) -> tuple[list[dict], str | None]: