diff --git a/python/sglang/srt/constrained/base_grammar_backend.py b/python/sglang/srt/constrained/base_grammar_backend.py
index d3be81ef7..dda3fab4f 100644
--- a/python/sglang/srt/constrained/base_grammar_backend.py
+++ b/python/sglang/srt/constrained/base_grammar_backend.py
@@ -224,13 +224,17 @@ def create_grammar_backend(
         eos_list = list(eos_token_ids) if eos_token_ids else None
 
         grammar_backend = XGrammarGrammarBackend(
-            tokenizer, vocab_size=vocab_size, model_eos_token_ids=eos_list
+            tokenizer,
+            vocab_size=vocab_size,
+            model_eos_token_ids=eos_list,
+            any_whitespace=not server_args.constrained_json_disable_any_whitespace,
         )
     elif name == "llguidance":
         from sglang.srt.constrained.llguidance_backend import GuidanceBackend
 
         grammar_backend = GuidanceBackend(
             tokenizer=tokenizer,
+            any_whitespace=not server_args.constrained_json_disable_any_whitespace,
             whitespace_pattern=server_args.constrained_json_whitespace_pattern,
         )
     elif name == "none":
diff --git a/python/sglang/srt/constrained/llguidance_backend.py b/python/sglang/srt/constrained/llguidance_backend.py
index 5e29c2524..dc34a353d 100644
--- a/python/sglang/srt/constrained/llguidance_backend.py
+++ b/python/sglang/srt/constrained/llguidance_backend.py
@@ -110,12 +110,14 @@ class GuidanceBackend(BaseGrammarBackend):
     def __init__(
         self,
         tokenizer,
+        any_whitespace: bool = True,
         whitespace_pattern: Optional[str] = None,
         n_vocab: Optional[int] = None,
     ):
         super().__init__()
 
         self.tokenizer = tokenizer
+        self.any_whitespace = any_whitespace
         self.whitespace_pattern = whitespace_pattern
         self.llguidance_tokenizer = from_tokenizer(self.tokenizer, n_vocab)
 
@@ -134,6 +136,7 @@ class GuidanceBackend(BaseGrammarBackend):
             serialized_grammar = LLMatcher.grammar_from_json_schema(
                 key_string,
                 defaults={
+                    "whitespace_flexible": self.any_whitespace,
                     "whitespace_pattern": self.whitespace_pattern,
                 },
             )
diff --git a/python/sglang/srt/constrained/outlines_backend.py b/python/sglang/srt/constrained/outlines_backend.py
index b54e34b3d..28831ab86 100644
--- a/python/sglang/srt/constrained/outlines_backend.py
+++ b/python/sglang/srt/constrained/outlines_backend.py
@@ -115,7 +115,7 @@ class OutlinesGrammarBackend(BaseGrammarBackend):
     def __init__(
         self,
         tokenizer,
-        whitespace_pattern: bool,
+        whitespace_pattern: str | None,
     ):
         super().__init__()
 
diff --git a/python/sglang/srt/constrained/xgrammar_backend.py b/python/sglang/srt/constrained/xgrammar_backend.py
index 3df788970..00b54baef 100644
--- a/python/sglang/srt/constrained/xgrammar_backend.py
+++ b/python/sglang/srt/constrained/xgrammar_backend.py
@@ -167,6 +167,7 @@ class XGrammarGrammarBackend(BaseGrammarBackend):
         tokenizer,
         vocab_size: int,
         model_eos_token_ids: Optional[List[int]] = None,
+        any_whitespace: bool = True,
     ):
         super().__init__()
 
@@ -188,6 +189,7 @@ class XGrammarGrammarBackend(BaseGrammarBackend):
         self.grammar_compiler = GrammarCompiler(tokenizer_info=tokenizer_info)
         self.vocab_size = vocab_size
         self.override_stop_tokens = override_stop_tokens
+        self.any_whitespace = any_whitespace
 
     def _from_context(
         self, ctx: CompiledGrammar, key_string: str, grammar_stats: GrammarStats
@@ -212,7 +214,9 @@ class XGrammarGrammarBackend(BaseGrammarBackend):
                 # Note: This builtin JSON grammar includes *all* valid JSON (including, for example, arrays at the root)
                 ctx = self.grammar_compiler.compile_builtin_json_grammar()
             else:
-                ctx = self.grammar_compiler.compile_json_schema(schema=key_string)
+                ctx = self.grammar_compiler.compile_json_schema(
+                    schema=key_string, any_whitespace=self.any_whitespace
+                )
 
         except (RuntimeError, json.decoder.JSONDecodeError) as e:
             logging.error(f"Hit invalid json_schema: {key_string=}, {e=}")
diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py
index c7e80a2b9..67d0a1807 100644
--- a/python/sglang/srt/server_args.py
+++ b/python/sglang/srt/server_args.py
@@ -227,6 +227,7 @@ class ServerArgs:
     stream_output: bool = False
     random_seed: Optional[int] = None
     constrained_json_whitespace_pattern: Optional[str] = None
+    constrained_json_disable_any_whitespace: bool = False
     watchdog_timeout: float = 300
     dist_timeout: Optional[int] = None  # timeout for torch.distributed
     download_dir: Optional[str] = None
@@ -1683,7 +1684,12 @@ class ServerArgs:
             "--constrained-json-whitespace-pattern",
             type=str,
             default=ServerArgs.constrained_json_whitespace_pattern,
-            help="(outlines backend only) Regex pattern for syntactic whitespaces allowed in JSON constrained output. For example, to allow the model generate consecutive whitespaces, set the pattern to [\n\t ]*",
+            help="(outlines and llguidance backends only) Regex pattern for syntactic whitespaces allowed in JSON constrained output. For example, to allow the model generate consecutive whitespaces, set the pattern to [\n\t ]*",
+        )
+        parser.add_argument(
+            "--constrained-json-disable-any-whitespace",
+            action="store_true",
+            help="(xgrammar and llguidance backends only) Enforce compact representation in JSON constrained output.",
         )
         parser.add_argument(
             "--watchdog-timeout",