diff --git a/docs/release-control/v6/internal/subsystems/ai-runtime.md b/docs/release-control/v6/internal/subsystems/ai-runtime.md
index c99c01d65..9017658c4 100644
--- a/docs/release-control/v6/internal/subsystems/ai-runtime.md
+++ b/docs/release-control/v6/internal/subsystems/ai-runtime.md
@@ -508,11 +508,14 @@ deriving an older display status from `workflowStatusHistory`.
    `packages/opencode/src/session/prompt.ts`. Pulse adapts that behavior for the
    drawer instead of implementing automatic cross-provider route switching.
    Primary interactive chat model resolution must use the explicit configured
-   chat route or a stable provider default without calling provider model
-   catalogs before the selected stream starts. Catalog-backed recommendation
-   belongs to settings/model-list and explicit route-recovery flows, not
-   `/api/ai/chat` first-response startup. Once the selected provider route
-   starts, transient pre-output transport failures may retry the same route and
+   chat route or, when no explicit chat route exists, a stable provider default
+   without calling provider model catalogs before the selected stream starts. If
+   an explicit chat route is unconfigured, retired, or recognized as a
+   specialized non-chat endpoint, `/api/ai/chat` must fail that selected route
+   visibly instead of substituting a same-provider default. Catalog-backed
+   recommendation belongs to settings/model-list and explicit route-recovery
+   flows, not `/api/ai/chat` first-response startup. Once the selected provider
+   route starts, transient pre-output transport failures may retry the same route and
    must surface a `provider_retry` workflow state with `attempt`,
    `max_attempts`, and `retry_after_ms` before sleeping. If same-route retry is
    exhausted before visible output, Pulse must emit a normal provider error and
diff --git a/internal/ai/chat/service.go b/internal/ai/chat/service.go
index a64eb4db4..d69977935 100644
--- a/internal/ai/chat/service.go
+++ b/internal/ai/chat/service.go
@@ -3688,7 +3688,7 @@ func (s *Service) createProvider() (providers.StreamingProvider, error) {
 	if chatModel == "" {
 		return nil, fmt.Errorf("no chat model configured")
 	}
-	if strings.TrimSpace(s.cfg.ChatModel) == "" || !modelresolution.IsModelUsableForChatWithConfig(s.cfg, s.cfg.GetChatModel()) {
+	if strings.TrimSpace(s.cfg.ChatModel) == "" {
 		s.cfg.ChatModel = chatModel
 	}
 
diff --git a/internal/ai/chat/service_test.go b/internal/ai/chat/service_test.go
index f0505eafe..1e64e3fdf 100644
--- a/internal/ai/chat/service_test.go
+++ b/internal/ai/chat/service_test.go
@@ -220,6 +220,20 @@ func TestService_CreateProvider(t *testing.T) {
 	}
 }
 
+func TestServiceCreateProviderRejectsUnusableConfiguredChatModelWithoutRewrite(t *testing.T) {
+	cfg := &config.AIConfig{
+		OpenAIAPIKey: "sk-test",
+		ChatModel:    "openai:gpt-realtime-2",
+	}
+	service := &Service{cfg: cfg}
+
+	provider, err := service.createProvider()
+	assert.Error(t, err)
+	assert.Nil(t, provider)
+	assert.Contains(t, err.Error(), "selected chat model route")
+	assert.Equal(t, "openai:gpt-realtime-2", cfg.ChatModel)
+}
+
 func TestService_ExecuteStream_Failures(t *testing.T) {
 	service := &Service{}
 	err := service.ExecuteStream(context.Background(), ExecuteRequest{}, nil)
diff --git a/internal/ai/model_resolution_test.go b/internal/ai/model_resolution_test.go
index 1c010ccef..a3d8930de 100644
--- a/internal/ai/model_resolution_test.go
+++ b/internal/ai/model_resolution_test.go
@@ -149,6 +149,24 @@ func TestResolveConfiguredChatProviderModel_SkipsSpecializedPreferredModel(t *te
 	}
 }
 
+func TestResolveConfiguredChatModel_RejectsSpecializedExplicitChatModel(t *testing.T) {
+	t.Parallel()
+
+	cfg := &config.AIConfig{
+		Enabled:      true,
+		OpenAIAPIKey: "sk-test",
+		ChatModel:    "openai:gpt-realtime-2",
+	}
+
+	got, err := ResolveConfiguredChatModel(context.Background(), cfg)
+	if err == nil {
+		t.Fatalf("ResolveConfiguredChatModel() error = nil, got model %q", got)
+	}
+	if !strings.Contains(err.Error(), "selected chat model route") || !strings.Contains(err.Error(), "gpt-realtime-2") {
+		t.Fatalf("ResolveConfiguredChatModel() error = %q, want selected chat model route error", err)
+	}
+}
+
 func TestResolveConfiguredModel_RejectsExplicitUnconfiguredRoute(t *testing.T) {
 	t.Parallel()
 
diff --git a/internal/ai/modelresolution/model_resolution.go b/internal/ai/modelresolution/model_resolution.go
index 363185253..e4d168dfc 100644
--- a/internal/ai/modelresolution/model_resolution.go
+++ b/internal/ai/modelresolution/model_resolution.go
@@ -46,11 +46,7 @@ func ResolveConfiguredChatModel(ctx context.Context, cfg *config.AIConfig) (stri
 		return explicit, nil
 	}
 	if explicit != "" {
-		provider, err := selectedModelProvider(cfg, explicit)
-		if err != nil {
-			return "", err
-		}
-		return ResolveConfiguredChatProviderModel(ctx, cfg, provider)
+		return "", selectedChatModelProviderError(cfg, explicit)
 	}
 
 	configuredProviders := cfg.GetConfiguredProviders()
@@ -75,11 +71,7 @@ func ResolveConfiguredChatModelOffline(cfg *config.AIConfig) (string, error) {
 		return explicit, nil
 	}
 	if explicit != "" {
-		provider, err := selectedModelProvider(cfg, explicit)
-		if err != nil {
-			return "", err
-		}
-		return ResolveConfiguredChatProviderModelOffline(cfg, provider)
+		return "", selectedChatModelProviderError(cfg, explicit)
 	}
 
 	configuredProviders := cfg.GetConfiguredProviders()
@@ -116,6 +108,13 @@ func selectedModelProviderError(cfg *config.AIConfig, model string) error {
 	return fmt.Errorf("selected model route %q is not usable with the current Pulse Assistant config", strings.TrimSpace(model))
 }
 
+func selectedChatModelProviderError(cfg *config.AIConfig, model string) error {
+	if _, err := selectedModelProvider(cfg, model); err != nil {
+		return err
+	}
+	return fmt.Errorf("selected chat model route %q is not usable for Assistant chat with the current Pulse Assistant config", strings.TrimSpace(model))
+}
+
 // ResolveConfiguredChatProviderModelOffline resolves a chat-suitable model for
 // a configured provider using only explicit config and stable provider defaults.
 func ResolveConfiguredChatProviderModelOffline(cfg *config.AIConfig, provider string) (string, error) {
diff --git a/internal/ai/modelresolution/model_resolution_offline_test.go b/internal/ai/modelresolution/model_resolution_offline_test.go
index b2782f771..d7d541d26 100644
--- a/internal/ai/modelresolution/model_resolution_offline_test.go
+++ b/internal/ai/modelresolution/model_resolution_offline_test.go
@@ -40,7 +40,7 @@ func TestResolveConfiguredChatModelOffline_UsesStableProviderDefault(t *testing.
 	}
 }
 
-func TestResolveConfiguredChatModelOffline_ReplacesSpecializedChatModelWithStableDefault(t *testing.T) {
+func TestResolveConfiguredChatModelOffline_RejectsSpecializedExplicitChatModel(t *testing.T) {
 	t.Parallel()
 
 	cfg := &config.AIConfig{
@@ -49,11 +49,11 @@ func TestResolveConfiguredChatModelOffline_ReplacesSpecializedChatModelWithStabl
 	}
 
 	got, err := ResolveConfiguredChatModelOffline(cfg)
-	if err != nil {
-		t.Fatalf("ResolveConfiguredChatModelOffline() error = %v", err)
+	if err == nil {
+		t.Fatalf("ResolveConfiguredChatModelOffline() error = nil, got model %q", got)
 	}
-	if want := config.DefaultModelForProvider(config.AIProviderOpenAI); got != want {
-		t.Fatalf("ResolveConfiguredChatModelOffline() = %q, want %q", got, want)
+	if !strings.Contains(err.Error(), "selected chat model route") || !strings.Contains(err.Error(), "gpt-realtime-2") {
+		t.Fatalf("ResolveConfiguredChatModelOffline() error = %q, want selected chat model route error", err)
 	}
 }