Reapply "don't require pulling stubs for cloud models"

This reverts commit 97d2f05a6d.

x/cmd/run.go

@@ -20,6 +20,7 @@ import (
 
 	"github.com/ollama/ollama/api"
 	internalcloud "github.com/ollama/ollama/internal/cloud"
+	"github.com/ollama/ollama/internal/modelref"
 	"github.com/ollama/ollama/progress"
 	"github.com/ollama/ollama/readline"
 	"github.com/ollama/ollama/types/model"
@@ -43,7 +44,7 @@ const (
 // isLocalModel checks if the model is running locally (not a cloud model).
 // TODO: Improve local/cloud model identification - could check model metadata
 func isLocalModel(modelName string) bool {
-	return !strings.HasSuffix(modelName, "-cloud")
+	return !modelref.HasExplicitCloudSource(modelName)
 }
 
 // isLocalServer checks if connecting to a local Ollama server.

x/cmd/run_test.go

@@ -22,12 +22,22 @@ func TestIsLocalModel(t *testing.T) {
 		},
 		{
 			name:      "cloud model",
-			modelName: "gpt-4-cloud",
+			modelName: "gpt-oss:latest-cloud",
+			expected:  false,
+		},
+		{
+			name:      "cloud model with :cloud suffix",
+			modelName: "gpt-oss:cloud",
 			expected:  false,
 		},
 		{
 			name:      "cloud model with version",
-			modelName: "claude-3-cloud",
+			modelName: "gpt-oss:20b-cloud",
+			expected:  false,
+		},
+		{
+			name:      "cloud model with version and :cloud suffix",
+			modelName: "gpt-oss:20b:cloud",
 			expected:  false,
 		},
 		{
@@ -134,7 +144,7 @@ func TestTruncateToolOutput(t *testing.T) {
 		{
 			name:          "long output cloud model - uses 10k limit",
 			output:        string(localLimitOutput), // 20k chars, under 10k token limit
-			modelName:     "gpt-4-cloud",
+			modelName:     "gpt-oss:latest-cloud",
 			host:          "",
 			shouldTrim:    false,
 			expectedLimit: defaultTokenLimit,
@@ -142,7 +152,7 @@ func TestTruncateToolOutput(t *testing.T) {
 		{
 			name:          "very long output cloud model - trimmed at 10k",
 			output:        string(defaultLimitOutput),
-			modelName:     "gpt-4-cloud",
+			modelName:     "gpt-oss:latest-cloud",
 			host:          "",
 			shouldTrim:    true,
 			expectedLimit: defaultTokenLimit,