diff --git a/frontend/types/gotypes.d.ts b/frontend/types/gotypes.d.ts
index 4658bc1af2..4f8f4a8473 100644
--- a/frontend/types/gotypes.d.ts
+++ b/frontend/types/gotypes.d.ts
@@ -23,6 +23,7 @@ declare global {
         "ai:apitype"?: string;
         "ai:model"?: string;
         "ai:thinkinglevel"?: string;
+        "ai:verbosity"?: string;
         "ai:endpoint"?: string;
         "ai:azureapiversion"?: string;
         "ai:apitoken"?: string;
diff --git a/pkg/aiusechat/openai/openai-convertmessage.go b/pkg/aiusechat/openai/openai-convertmessage.go
index fd0df85486..3accb354fa 100644
--- a/pkg/aiusechat/openai/openai-convertmessage.go
+++ b/pkg/aiusechat/openai/openai-convertmessage.go
@@ -22,6 +22,8 @@ import (
 const (
 	OpenAIDefaultAPIVersion = "2024-12-31"
 	OpenAIDefaultMaxTokens  = 4096
+	// "medium" verbosity is more widely supported across models than "low"
+	OpenAIDefaultVerbosity = "medium"
 )
 
 // convertContentBlockToParts converts a single content block to UIMessageParts
@@ -190,10 +192,11 @@ func debugPrintReq(req *OpenAIRequest, endpoint string) {
 func buildOpenAIHTTPRequest(ctx context.Context, inputs []any, chatOpts uctypes.WaveChatOpts, cont *uctypes.WaveContinueResponse) (*http.Request, error) {
 	opts := chatOpts.Config
 
-	// If continuing from premium rate limit, downgrade to default model and low thinking
+	// If continuing from premium rate limit, downgrade to default model and medium thinking
+	// (medium is more widely supported than low across different models)
 	if cont != nil && cont.ContinueFromKind == uctypes.StopKindPremiumRateLimit {
 		opts.Model = uctypes.DefaultOpenAIModel
-		opts.ThinkingLevel = uctypes.ThinkingLevelLow
+		opts.ThinkingLevel = uctypes.ThinkingLevelMedium
 	}
 
 	if opts.Model == "" {
@@ -229,13 +232,18 @@ func buildOpenAIHTTPRequest(ctx context.Context, inputs []any, chatOpts uctypes.
 	}
 
 	// Build request body
+	// Use configured verbosity, or fall back to default constant
+	verbosity := opts.Verbosity
+	if verbosity == "" {
+		verbosity = OpenAIDefaultVerbosity
+	}
 	reqBody := &OpenAIRequest{
 		Model:           opts.Model,
 		Input:           inputs,
 		Stream:          true,
 		StreamOptions:   &StreamOptionsType{IncludeObfuscation: false},
 		MaxOutputTokens: maxTokens,
-		Text:            &TextType{Verbosity: "low"},
+		Text:            &TextType{Verbosity: verbosity},
 	}
 
 	// Add system prompt as instructions if provided
@@ -264,10 +272,10 @@ func buildOpenAIHTTPRequest(ctx context.Context, inputs []any, chatOpts uctypes.
 		reqBody.Tools = append(reqBody.Tools, webSearchTool)
 	}
 
-	// Set reasoning based on thinking level
+	// Set reasoning based on thinking level from config
 	if opts.ThinkingLevel != "" {
 		reqBody.Reasoning = &ReasoningType{
-			Effort: opts.ThinkingLevel, // low, medium, high map directly
+			Effort: opts.ThinkingLevel,
 		}
 		if opts.Model == "gpt-5" || opts.Model == "gpt-5.1" {
 			reqBody.Reasoning.Summary = "auto"
diff --git a/pkg/aiusechat/uctypes/uctypes.go b/pkg/aiusechat/uctypes/uctypes.go
index b857f141bd..fac8cb7f2a 100644
--- a/pkg/aiusechat/uctypes/uctypes.go
+++ b/pkg/aiusechat/uctypes/uctypes.go
@@ -278,6 +278,7 @@ type AIOptsType struct {
 	MaxTokens     int      `json:"maxtokens,omitempty"`
 	TimeoutMs     int      `json:"timeoutms,omitempty"`
 	ThinkingLevel string   `json:"thinkinglevel,omitempty"` // ThinkingLevelLow, ThinkingLevelMedium, or ThinkingLevelHigh
+	Verbosity     string   `json:"verbosity,omitempty"`     // Text verbosity level (OpenAI Responses API only, ignored by other backends)
 	AIMode        string   `json:"aimode,omitempty"`
 	Capabilities  []string `json:"capabilities,omitempty"`
 	WaveAIPremium bool     `json:"waveaipremium,omitempty"`
diff --git a/pkg/aiusechat/usechat.go b/pkg/aiusechat/usechat.go
index 8d8fcf6446..46ec030634 100644
--- a/pkg/aiusechat/usechat.go
+++ b/pkg/aiusechat/usechat.go
@@ -110,12 +110,17 @@ func getWaveAISettings(premium bool, builderMode bool, rtInfo waveobj.ObjRTInfo,
 	if thinkingLevel == "" {
 		thinkingLevel = uctypes.ThinkingLevelMedium
 	}
+	verbosity := config.Verbosity
+	if verbosity == "" {
+		verbosity = uctypes.ThinkingLevelMedium // default to medium
+	}
 	opts := &uctypes.AIOptsType{
 		Provider:      config.Provider,
 		APIType:       config.APIType,
 		Model:         config.Model,
 		MaxTokens:     maxTokens,
 		ThinkingLevel: thinkingLevel,
+		Verbosity:     verbosity,
 		AIMode:        aiMode,
 		Endpoint:      baseUrl,
 		Capabilities:  config.Capabilities,
diff --git a/pkg/wconfig/settingsconfig.go b/pkg/wconfig/settingsconfig.go
index 50a1da2474..efeedfc6db 100644
--- a/pkg/wconfig/settingsconfig.go
+++ b/pkg/wconfig/settingsconfig.go
@@ -273,6 +273,7 @@ type AIModeConfigType struct {
 	APIType            string   `json:"ai:apitype,omitempty" jsonschema:"enum=google-gemini,enum=openai-responses,enum=openai-chat"`
 	Model              string   `json:"ai:model,omitempty"`
 	ThinkingLevel      string   `json:"ai:thinkinglevel,omitempty" jsonschema:"enum=low,enum=medium,enum=high"`
+	Verbosity          string   `json:"ai:verbosity,omitempty" jsonschema:"enum=low,enum=medium,enum=high,description=Text verbosity level (OpenAI Responses API only)"`
 	Endpoint           string   `json:"ai:endpoint,omitempty"`
 	AzureAPIVersion    string   `json:"ai:azureapiversion,omitempty"`
 	APIToken           string   `json:"ai:apitoken,omitempty"`
diff --git a/schema/waveai.json b/schema/waveai.json
index 1612cac88d..1c3a186beb 100644
--- a/schema/waveai.json
+++ b/schema/waveai.json
@@ -46,6 +46,15 @@
             "high"
           ]
         },
+        "ai:verbosity": {
+          "type": "string",
+          "enum": [
+            "low",
+            "medium",
+            "high"
+          ],
+          "description": "Text verbosity level (OpenAI Responses API only)"
+        },
         "ai:endpoint": {
           "type": "string"
         },