diff --git a/studio/backend/core/inference/external_provider.py b/studio/backend/core/inference/external_provider.py
index df06098156..86ea6ea800 100644
--- a/studio/backend/core/inference/external_provider.py
+++ b/studio/backend/core/inference/external_provider.py
@@ -27,6 +27,52 @@ import structlog
 logger = structlog.get_logger(__name__)
 
 
+def _normalize_stop_for_provider(
+    stop: Optional[Union[str, list[str]]],
+    provider_info: dict[str, Any],
+) -> Optional[Union[str, list[str]]]:
+    """Apply per-provider stop_max / stop_max_bytes caps and dedup.
+
+    Returns None when nothing survives the filter so callers can omit
+    the field. Single strings are returned verbatim when they fit.
+    """
+    if not stop:
+        return None
+
+    stop_max = int(provider_info.get("stop_max", 16))
+    stop_max_bytes_raw = provider_info.get("stop_max_bytes")
+    stop_max_bytes = (
+        int(stop_max_bytes_raw) if stop_max_bytes_raw is not None else None
+    )
+
+    def allowed(s: str) -> bool:
+        if not s:
+            return False
+        if stop_max_bytes is not None and len(s.encode("utf-8")) > stop_max_bytes:
+            logger.warning(
+                "dropping stop sequence longer than %d bytes",
+                stop_max_bytes,
+            )
+            return False
+        return True
+
+    if isinstance(stop, str):
+        return stop if allowed(stop) else None
+    if isinstance(stop, list):
+        sequences = list(
+            dict.fromkeys(s for s in stop if isinstance(s, str) and allowed(s))
+        )
+        if len(sequences) > stop_max:
+            logger.warning(
+                "stop sequences truncated to %d entries (received %d)",
+                stop_max,
+                len(sequences),
+            )
+            sequences = sequences[:stop_max]
+        return sequences or None
+    return None
+
+
 # Claude 4.7 (Opus/Sonnet/Haiku) removed temperature, top_p, and top_k —
 # the API returns 400 "<param> is deprecated for this model" if any of
 # them is set to a non-default value. The "Sampling parameters removed"
@@ -472,25 +518,9 @@ class ExternalProviderClient:
             # Mistral renames `seed` to `random_seed` on /v1/chat/completions.
             seed_field = provider_info.get("seed_field", "seed")
             body[seed_field] = seed
-        if stop:
-            # Stop cap is provider-specific. OpenAI Chat = 4, Anthropic
-            # = 16 (client guard), DeepSeek = 16, others = 16 by default.
-            stop_max = int(provider_info.get("stop_max", 16))
-            if isinstance(stop, str):
-                body["stop"] = stop
-            elif isinstance(stop, list):
-                sequences = list(
-                    dict.fromkeys(s for s in stop if isinstance(s, str) and s)
-                )
-                if len(sequences) > stop_max:
-                    logger.warning(
-                        "stop sequences truncated to %d entries (received %d)",
-                        stop_max,
-                        len(sequences),
-                    )
-                    body["stop"] = sequences[:stop_max]
-                elif sequences:
-                    body["stop"] = sequences
+        normalized_stop = _normalize_stop_for_provider(stop, provider_info)
+        if normalized_stop:
+            body["stop"] = normalized_stop
         if service_tier is not None:
             body["service_tier"] = service_tier
         if parallel_tool_calls is not None:
@@ -908,23 +938,9 @@ class ExternalProviderClient:
         if seed is not None:
             seed_field = provider_info.get("seed_field", "seed")
             body[seed_field] = seed
-        if stop:
-            stop_max = int(provider_info.get("stop_max", 16))
-            if isinstance(stop, str):
-                body["stop"] = stop
-            elif isinstance(stop, list):
-                sequences = list(
-                    dict.fromkeys(s for s in stop if isinstance(s, str) and s)
-                )
-                if len(sequences) > stop_max:
-                    logger.warning(
-                        "stop sequences truncated to %d entries (received %d)",
-                        stop_max,
-                        len(sequences),
-                    )
-                    body["stop"] = sequences[:stop_max]
-                elif sequences:
-                    body["stop"] = sequences
+        normalized_stop = _normalize_stop_for_provider(stop, provider_info)
+        if normalized_stop:
+            body["stop"] = normalized_stop
         if parallel_tool_calls is not None:
             body["parallel_tool_calls"] = parallel_tool_calls
 
diff --git a/studio/backend/core/inference/providers.py b/studio/backend/core/inference/providers.py
index 4a47e90850..6701b27ba9 100644
--- a/studio/backend/core/inference/providers.py
+++ b/studio/backend/core/inference/providers.py
@@ -165,6 +165,7 @@ PROVIDER_REGISTRY: dict[str, dict[str, Any]] = {
         # Kimi accepts at most 5 stop strings (each <= 32 bytes) per
         # https://platform.kimi.ai/docs/api/chat
         "stop_max": 5,
+        "stop_max_bytes": 32,
     },
     "qwen": {
         "display_name": "Qwen",
diff --git a/studio/backend/tests/test_sampling_params_routing.py b/studio/backend/tests/test_sampling_params_routing.py
index 012c2266f2..99cb3a8993 100644
--- a/studio/backend/tests/test_sampling_params_routing.py
+++ b/studio/backend/tests/test_sampling_params_routing.py
@@ -607,6 +607,61 @@ def test_kimi_web_search_uses_kimi_stop_cap_5(monkeypatch):
     assert body["stop"] == ["S0", "S1", "S2", "S3", "S4"], body
 
 
+def test_kimi_drops_stop_strings_over_32_bytes(monkeypatch):
+    """Kimi limits each stop string to <= 32 bytes per
+    https://platform.kimi.ai/docs/api/chat. Drop overlong entries
+    client-side so a stale UI cannot 400 the request."""
+    captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
+
+    async def run():
+        client = ExternalProviderClient(
+            provider_type = "kimi",
+            base_url = "https://api.moonshot.ai/v1",
+            api_key = "kimi-test",
+        )
+        async for _ in client.stream_chat_completion(
+            messages = [{"role": "user", "content": "hi"}],
+            model = "kimi-k2.6",
+            temperature = 1.0,
+            top_p = 1.0,
+            max_tokens = 256,
+            stop = ["END", "x" * 33, "DONE"],
+        ):
+            pass
+        await client.close()
+
+    _drive(run())
+    body = captured["body"]
+    assert body.get("stop") == ["END", "DONE"], body
+
+
+def test_kimi_web_search_drops_stop_strings_over_32_bytes(monkeypatch):
+    """Same byte cap applies to the Kimi web-search bypass."""
+    captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())
+
+    async def run():
+        client = ExternalProviderClient(
+            provider_type = "kimi",
+            base_url = "https://api.moonshot.ai/v1",
+            api_key = "kimi-test",
+        )
+        async for _ in client.stream_chat_completion(
+            messages = [{"role": "user", "content": "hi"}],
+            model = "kimi-k2.6",
+            temperature = 1.0,
+            top_p = 1.0,
+            max_tokens = 256,
+            enabled_tools = ["web_search"],
+            stop = ["END", "x" * 40, "DONE"],
+        ):
+            pass
+        await client.close()
+
+    _drive(run())
+    body = captured["body"]
+    assert body.get("stop") == ["END", "DONE"], body
+
+
 def test_kimi_default_path_uses_kimi_stop_cap_5(monkeypatch):
     """The normal Kimi path must also honour the documented 5-cap."""
     captured = _install_mock(monkeypatch, sse_payload = _oai_done_payload())