diff --git a/agent/model_metadata.py b/agent/model_metadata.py
index 821a7ebe0..98a559bfd 100644
--- a/agent/model_metadata.py
+++ b/agent/model_metadata.py
@@ -319,7 +319,10 @@ DEFAULT_CONTEXT_LENGTHS = {
     "grok-3": 131072,           # grok-3, grok-3-mini, grok-3-fast, grok-3-mini-fast
     "grok-2": 131072,           # grok-2, grok-2-1212, grok-2-latest
     "grok": 131072,             # catch-all (grok-beta, unknown grok-*)
-    # Kimi
+    # Kimi — K3 ships with a 1M context window (verified: platform.kimi.ai/docs/overview).
+    # Longest-key-first substring matching ensures "kimi-k3" resolves to 1M
+    # while older/unknown Kimi models still hit the generic 256K fallback.
+    "kimi-k3": 1_000_000,
     "kimi": 262144,
     # Upstage Solar — api.upstage.ai/v1/models does not return context_length,
     # so these fallbacks keep token budgeting / compression from probing down
diff --git a/tests/agent/test_model_metadata.py b/tests/agent/test_model_metadata.py
index ddd49834d..1789aa559 100644
--- a/tests/agent/test_model_metadata.py
+++ b/tests/agent/test_model_metadata.py
@@ -332,6 +332,31 @@ class TestDefaultContextLengths:
             assert get_model_context_length("glm-5") == 202752
             assert get_model_context_length("glm-5.1") == 202752
 
+    def test_kimi_k3_context_1m(self):
+        """Kimi K3 must resolve to 1M, not the generic Kimi fallback of 256K.
+
+        Context window verified against platform.kimi.ai/docs/overview
+        (2026-07). Kimi K3 is the current flagship model with a 1M-token
+        context window.
+        """
+        from agent.model_metadata import get_model_context_length
+        from unittest.mock import patch as mock_patch
+
+        assert DEFAULT_CONTEXT_LENGTHS["kimi-k3"] == 1_000_000
+        assert DEFAULT_CONTEXT_LENGTHS["kimi"] == 262144
+
+        with mock_patch("agent.model_metadata.fetch_model_metadata", return_value={}), \
+             mock_patch("agent.model_metadata.fetch_endpoint_model_metadata", return_value={}), \
+             mock_patch("agent.model_metadata.get_cached_context_length", return_value=None):
+            # Kimi K3 (1M) must NOT fall through to the generic 256K entry
+            assert get_model_context_length("kimi-k3") == 1_000_000
+            # Vendor-prefixed forms (kimi provider, openrouter)
+            assert get_model_context_length("kimi/kimi-k3") == 1_000_000
+            assert get_model_context_length("moonshotai/kimi-k3") == 1_000_000
+            # Older/unknown Kimi models still resolve to 256K fallback
+            assert get_model_context_length("kimi-k2.6") == 262144
+            assert get_model_context_length("kimi-k2") == 262144
+
     def test_openrouter_live_metadata_beats_hardcoded_catchall(self):
         """OpenRouter-routed slugs resolve via the live OR catalog before the
         hardcoded family catch-all.