diff --git a/backend/internal/service/finance_price_tier_test.go b/backend/internal/service/finance_price_tier_test.go
index 4b3930cbb..c95a9f032 100644
--- a/backend/internal/service/finance_price_tier_test.go
+++ b/backend/internal/service/finance_price_tier_test.go
@@ -1,7 +1,10 @@
 package service
 
 import (
+	"fmt"
+	"strings"
 	"testing"
+	"time"
 
 	"infinite-canvas/backend/internal/model"
 	"infinite-canvas/backend/internal/repository"
@@ -10,6 +13,72 @@ import (
 	"gorm.io/gorm"
 )
 
+func TestImageSpecificationQuoteAgreesWithTaskBilling(t *testing.T) {
+	db, err := gorm.Open(sqlite.Open(":memory:"), &gorm.Config{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := db.AutoMigrate(&model.SystemSetting{}, &model.ChannelModel{}, &model.ChannelModelPriceTier{}); err != nil {
+		t.Fatal(err)
+	}
+	channelModel := model.ChannelModel{
+		ID: "image-channel-model", ChannelID: "image-channel", ModelKey: "image-model", Capability: "image",
+		Protocol: model.ChannelInterfaceOpenAIImage, Enabled: true, PriceConfigured: true,
+	}
+	for operationIndex, operation := range []string{"text_to_image", "image_to_image"} {
+		for index, quality := range []string{"1k", "2k", "4k"} {
+			selector := fmt.Sprintf(`{"operation":%q,"quality":%q}`, operation, quality)
+			channelModel.PriceTiers = append(channelModel.PriceTiers, model.ChannelModelPriceTier{
+				ID: operation + "-" + quality, ChannelModelID: channelModel.ID, SelectorKey: selector, SelectorJSON: selector,
+				Resolution: "*", BillingMode: "fixed_request", UnitPriceMicrocredits: int64(index+1+operationIndex*3) * 1_000_000,
+				Enabled: true, PriceConfigured: true,
+			})
+		}
+	}
+	if err := db.Create(&channelModel).Error; err != nil {
+		t.Fatal(err)
+	}
+	if err := db.Create(&channelModel.PriceTiers).Error; err != nil {
+		t.Fatal(err)
+	}
+	spec := CapabilitySpec{Version: 1, Capability: "image", Inputs: map[string]InputConstraint{"image": {Max: 9}}, Options: map[string]OptionConstraint{"quality": {Values: []any{"1k", "2k", "4k"}}}}
+	cached := cachedLogicalModel{
+		Model:       model.LogicalModel{ID: "logical-image", PricePolicy: "channel", Capability: "image"},
+		ProductSpec: spec, Defaults: map[string]any{"quality": "2k"},
+		Routes: []cachedLogicalRoute{{Route: model.LogicalModelRoute{ID: "image-route", Enabled: true, Weight: 1}, CapabilitySpec: spec, ChannelModel: channelModel}},
+	}
+	svc := &Service{repo: repository.New(db), routeCatalogTTL: time.Hour, routeCatalog: &routeCatalogSnapshot{LoadedAt: time.Now(), Models: map[string]cachedLogicalModel{"logical-image": cached}}}
+	for _, tier := range channelModel.PriceTiers {
+		selector := model.DecodeSKUSelector(tier.SelectorJSON)
+		imageCount := 0
+		if selector["operation"] == "image_to_image" {
+			imageCount = 1
+		}
+		input := map[string]any{"mode": "image", "referenceImages": make([]any, imageCount), "config": map[string]any{
+			"channelId": channelModel.ChannelID, "model": channelModel.ModelKey, "quality": strings.ToUpper(selector["quality"]),
+		}}
+		intent := ModelRequestIntentFromTaskInput(input, "canvas_image", "image")
+		quote, err := svc.QuoteLogicalModel("logical-image", intent)
+		if err != nil {
+			t.Fatalf("quote %s: %v", tier.ID, err)
+		}
+		order, err := svc.taskBillingOrder("user", &model.Task{ID: "task", Type: "canvas_image", Operation: "image"}, input)
+		if err != nil {
+			t.Fatalf("billing %s: %v", tier.ID, err)
+		}
+		if quote.AmountMicrocredits != tier.UnitPriceMicrocredits || order.AmountMicrocredits != quote.AmountMicrocredits || order.PriceTierID != tier.ID {
+			t.Fatalf("tier %s: quote=%#v order=%#v", tier.ID, quote, order)
+		}
+	}
+	quote, err := svc.QuoteLogicalModel("logical-image", ModelRequestIntent{Capability: "image"})
+	if err != nil || quote.AmountMicrocredits != 2_000_000 {
+		t.Fatalf("default resolution quote=%#v error=%v", quote, err)
+	}
+	if _, err := svc.QuoteLogicalModel("logical-image", ModelRequestIntent{Capability: "image", Options: map[string]any{"quality": "8k"}}); err == nil {
+		t.Fatal("unconfigured resolution must not be quoted")
+	}
+}
+
 func TestTaskBillingOrderMatchesSystemImagePriceTierFromRequestedSpec(t *testing.T) {
 	db, err := gorm.Open(sqlite.Open("file:"+newID()+"?mode=memory&cache=shared"), &gorm.Config{})
 	if err != nil {
diff --git a/backend/internal/service/model_router.go b/backend/internal/service/model_router.go
index 64d1c9828..0be46a285 100644
--- a/backend/internal/service/model_router.go
+++ b/backend/internal/service/model_router.go
@@ -85,7 +85,14 @@ func ModelRequestIntentFromTaskInput(input map[string]any, taskType string, oper
 }
 
 func normalizeModelRequestOption(name string, value any) any {
-	if canonicalCapabilityOptionName(name) != "vquality" {
+	canonicalName := canonicalCapabilityOptionName(name)
+	if canonicalName == "quality" || canonicalName == "size" {
+		if text, ok := value.(string); ok {
+			return strings.ToLower(strings.TrimSpace(text))
+		}
+		return value
+	}
+	if canonicalName != "vquality" {
 		return value
 	}
 	resolution, ok := value.(string)
@@ -702,7 +709,18 @@ func skuSelectorForIntent(intent ModelRequestIntent) map[string]string {
 			selector["videoSeconds"] = strconv.Itoa(seconds)
 		}
 	case "image":
+		if intent.Inputs["image"] > 0 {
+			selector["operation"] = "image_to_image"
+		} else {
+			selector["operation"] = "text_to_image"
+		}
+		if quality := normalizeImagePriceQuality(fmt.Sprint(intent.Options["quality"]), fmt.Sprint(intent.Options["size"])); quality != "" {
+			selector["quality"] = quality
+		}
 		for _, key := range []string{"quality", "size"} {
+			if key == "quality" && selector["quality"] != "" {
+				continue
+			}
 			if value := strings.ToLower(strings.TrimSpace(fmt.Sprint(intent.Options[key]))); value != "" && value != "auto" && value != "any" {
 				selector[key] = value
 			}
@@ -711,6 +729,33 @@ func skuSelectorForIntent(intent ModelRequestIntent) map[string]string {
 	return selector
 }
 
+func normalizeImagePriceQuality(rawQuality string, rawSize string) string {
+	quality := strings.ToLower(strings.TrimSpace(rawQuality))
+	if quality != "" && quality != "auto" && quality != "any" {
+		return quality
+	}
+	parts := strings.Split(strings.ToLower(strings.TrimSpace(rawSize)), "x")
+	if len(parts) != 2 {
+		return ""
+	}
+	width, widthErr := strconv.ParseInt(strings.TrimSpace(parts[0]), 10, 64)
+	height, heightErr := strconv.ParseInt(strings.TrimSpace(parts[1]), 10, 64)
+	if widthErr != nil || heightErr != nil || width <= 0 || height <= 0 || width > (1<<32)/height {
+		return ""
+	}
+	pixels := width * height
+	switch {
+	case pixels <= 2_000_000:
+		return "1k"
+	case pixels <= 4_300_000:
+		return "2k"
+	case pixels <= 8_294_400:
+		return "4k"
+	default:
+		return ""
+	}
+}
+
 func skuSelectorForTier(tier model.ChannelModelPriceTier) map[string]string {
 	selector := model.DecodeSKUSelector(tier.SelectorJSON)
 	if len(selector) == 0 {
diff --git a/backend/internal/service/model_router_test.go b/backend/internal/service/model_router_test.go
index e076f1e74..cfe50c785 100644
--- a/backend/internal/service/model_router_test.go
+++ b/backend/internal/service/model_router_test.go
@@ -1,11 +1,43 @@
 package service
 
 import (
+	"fmt"
 	"testing"
 
 	"infinite-canvas/backend/internal/model"
 )
 
+func TestImagePriceTiersMatchResolutionAndActualReferences(t *testing.T) {
+	channelModel := model.ChannelModel{}
+	for _, operation := range []string{"text_to_image", "image_to_image"} {
+		for _, quality := range []string{"1k", "2k", "4k"} {
+			channelModel.PriceTiers = append(channelModel.PriceTiers, model.ChannelModelPriceTier{
+				ID: operation + "-" + quality, SelectorJSON: fmt.Sprintf(`{"operation":%q,"quality":%q}`, operation, quality), Enabled: true, PriceConfigured: true,
+			})
+		}
+	}
+	for _, operation := range []string{"", "image", "text_to_image", "image_to_image"} {
+		for _, quality := range []string{"1K", "2K", "4K"} {
+			for _, imageCount := range []int{0, 1, 3} {
+				intent := ModelRequestIntentFromTaskInput(map[string]any{
+					"mode": "image", "referenceImages": make([]any, imageCount), "config": map[string]any{"quality": quality},
+				}, "canvas_image", operation)
+				wantOperation := "text_to_image"
+				if imageCount > 0 {
+					wantOperation = "image_to_image"
+				}
+				tier := channelModelPriceTierForIntent(channelModel, intent)
+				if tier == nil || tier.ID != wantOperation+"-"+skuSelectorForIntent(intent)["quality"] {
+					t.Fatalf("operation=%q quality=%q imageCount=%d: tier=%#v", operation, quality, imageCount, tier)
+				}
+			}
+		}
+	}
+	if tier := channelModelPriceTierForIntent(channelModel, ModelRequestIntent{Capability: "image", Options: map[string]any{"quality": "8k"}}); tier != nil {
+		t.Fatalf("unconfigured resolution matched tier: %#v", tier)
+	}
+}
+
 func TestModelRequestIntentNormalizesVideoResolution(t *testing.T) {
 	input := map[string]any{
 		"mode":   "video",
@@ -17,6 +49,36 @@ func TestModelRequestIntentNormalizesVideoResolution(t *testing.T) {
 	}
 }
 
+func TestModelRequestIntentNormalizesImageSpecificationValues(t *testing.T) {
+	input := map[string]any{
+		"mode":   "image",
+		"config": map[string]any{"quality": "1K", "size": "3:2"},
+	}
+	intent := ModelRequestIntentFromTaskInput(input, "canvas_image", "image")
+	if got := intent.Options["quality"]; got != "1k" {
+		t.Fatalf("quality = %#v, want 1k", got)
+	}
+	if got := intent.Options["size"]; got != "3:2" {
+		t.Fatalf("size = %#v, want 3:2", got)
+	}
+}
+
+func TestSKUSelectorInfersImageResolutionFromSizeWhenQualityIsAutomatic(t *testing.T) {
+	for _, test := range []struct {
+		size string
+		want string
+	}{
+		{size: "1024x1024", want: "1k"},
+		{size: "2048x2048", want: "2k"},
+		{size: "2880x2880", want: "4k"},
+	} {
+		selector := skuSelectorForIntent(ModelRequestIntent{Capability: "image", Options: map[string]any{"quality": "auto", "size": test.size}})
+		if selector["quality"] != test.want {
+			t.Fatalf("size %s quality = %q, want %q", test.size, selector["quality"], test.want)
+		}
+	}
+}
+
 func TestSKUSelectorIncludesVideoReferenceImageCount(t *testing.T) {
 	selector := skuSelectorForIntent(ModelRequestIntent{Capability: "video", Inputs: map[string]int{"image": 5}, Options: map[string]any{"vquality": "720p"}})
 	if selector["imageCount"] != "5" || selector["vquality"] != "720p" {
diff --git a/docs/content/docs/progress/pending-test.mdx b/docs/content/docs/progress/pending-test.mdx
index e276dd564..42da8c3cc 100644
--- a/docs/content/docs/progress/pending-test.mdx
+++ b/docs/content/docs/progress/pending-test.mdx
@@ -5,6 +5,13 @@ description: 当前项目已实现功能的手工验证清单与未确认边界
 
 # 待测试
 
+## 图片规格价格
+
+- [ ] 仅配置 1K、2K、4K 三档价格（无统一兜底），画布逐一切换分辨率，显示与对应档位一致的积分价格。
+- [ ] 分别配置文生图、图生图规格价格；添加、移除参考图或角色参考后，报价与实际生成命中对应档位。
+- [ ] 新建图片节点时，报价使用节点模型默认规格，而不是全局旧参数；刷新后再次核对。
+- [ ] 配置统一价格时仍正常兜底；未配置且没有兜底的规格不能被计为其他档位价格。
+
 ## 编辑器（M2 命令状态机）
 
 - [x] 命令协议黄金文件：`editor-commands` 8 个黄金 op（addClip/moveClip/trimClip/splitClip/removeClip/setClipProperty/addSubtitle/removeSubtitle）全量快照比对通过；未知 op、非法 payload 一律抛错（fail-closed）且输入不可变。
diff --git a/web/src/components/canvas/canvas-node-prompt-panel.tsx b/web/src/components/canvas/canvas-node-prompt-panel.tsx
index ca00c3cdc..2ec930f1a 100644
--- a/web/src/components/canvas/canvas-node-prompt-panel.tsx
+++ b/web/src/components/canvas/canvas-node-prompt-panel.tsx
@@ -9,6 +9,7 @@ import { CreditSymbol, requestCreditCost } from "@/constant/credits";
 import { canvasThemes } from "@/lib/canvas-theme";
 import { modelQuoteRequest } from "@/lib/model-pricing";
 import { normalizeVideoDuration, normalizeVideoResolution } from "@/lib/video-generation-options";
+import { buildImageResolutionOptions, imageResolutionOption } from "@/lib/image-resolution-tiers";
 import { modelRequestOptions, resolveCompatibleModel, resolveModelGenerationDefaults, defaultImageParamsForModel, type ModelRequirements } from "@/lib/model-selection";
 import { navigateToSettings } from "@/lib/settings-navigation";
 import { useThemeStore } from "@/stores/use-theme-store";
@@ -107,6 +108,11 @@ export function CanvasNodePromptPanel({ node, isRunning, onPromptChange, onConfi
         }, mode),
     };
     const config = buildNodeConfig(globalConfig, node, mode, requirements);
+    const resolvedRequirements: ModelRequirements = {
+        ...requirements,
+        options: modelRequestOptions(config, mode),
+        videoSeconds: mode === "video" ? config.videoSeconds : undefined,
+    };
     const promptOptimizerProvider = useMemo(() => {
         if (!promptOptimizerEnabled || !promptOptimizerInstallation || !promptOptimizerPlugin.createPromptOptimizer) return null;
         return promptOptimizerPlugin.createPromptOptimizer(createPluginHostContext(promptOptimizerPlugin, promptOptimizerInstallation, globalConfig));
@@ -121,9 +127,9 @@ export function CanvasNodePromptPanel({ node, isRunning, onPromptChange, onConfi
         seconds: mode === "video" ? config.videoSeconds : 1,
         capability: mode,
         config,
-        requirements,
+        requirements: resolvedRequirements,
     });
-    const quoteRequest = modelQuoteRequest(config, config.model, mode, requirements);
+    const quoteRequest = modelQuoteRequest(config, config.model, mode, resolvedRequirements);
     const quoteRequestKey = JSON.stringify(quoteRequest || null);
     const [quotedCredits, setQuotedCredits] = useState<number | null>(null);
     const credits = quotedCredits ?? configuredCredits;
@@ -345,7 +351,7 @@ export function CanvasNodePromptPanel({ node, isRunning, onPromptChange, onConfi
                         value={config.model}
                         onChange={(model) => onConfigChange(node.id, mode === "image" ? { model, ...defaultImageParamsForModel(config, model) } : { model })}
                         capability={mode}
-                        requirements={requirements}
+                        requirements={resolvedRequirements}
                         onMissingConfig={() => navigateToSettings({ continueCreation: true })}
                         showSelectedPrice={false}
                         variant="creation"
@@ -734,10 +740,11 @@ function buildNodeConfig(globalConfig: AiConfig, node: CanvasNodeData, mode: Can
             videoWatermark: globalConfig.videoWatermark || defaultConfig.videoWatermark,
         },
     );
+    const imageQuality = mode === "image" ? resolvePricedImageQuality({ ...globalConfig, size: defaults.size ?? globalConfig.size ?? defaultConfig.size }, model, defaults.quality || globalConfig.quality || defaultConfig.quality) : undefined;
     return {
         ...globalConfig,
         model,
-        quality: defaults.quality || globalConfig.quality || defaultConfig.quality,
+        quality: imageQuality || defaults.quality || globalConfig.quality || defaultConfig.quality,
         size: defaults.size ?? globalConfig.size ?? defaultConfig.size,
         transparentBackground: defaults.transparentBackground || "false",
         videoSeconds: defaults.videoSeconds || normalizeVideoDuration(globalConfig.videoSeconds || defaultConfig.videoSeconds),
@@ -752,6 +759,19 @@ function buildNodeConfig(globalConfig: AiConfig, node: CanvasNodeData, mode: Can
     };
 }
 
+export function resolvePricedImageQuality(config: AiConfig, model: string, quality: string) {
+    if (quality && quality !== "auto") return quality;
+    const channel = resolveModelChannel(config, model);
+    const sizeTier = imageResolutionOption(buildImageResolutionOptions([config.size]), config.size)?.tier;
+    if (sizeTier) return sizeTier;
+    const cost = channel.modelCosts?.find((item) => item.model === modelOptionName(model));
+    const tier = cost?.logicalPriceTiers?.find((item) => {
+        const value = item.selector?.quality || "";
+        return value && value !== "*";
+    });
+    return tier?.selector?.quality || quality;
+}
+
 function promptPlaceholder(mode: CanvasNodeGenerationMode, hasImageContent: boolean, hasTextContent: boolean) {
     if (mode === "video") return "描述要生成的视频内容";
     if (mode === "audio") return "描述要生成的音频内容";
diff --git a/web/src/lib/model-pricing.ts b/web/src/lib/model-pricing.ts
index 92bfab8af..1e32d798c 100644
--- a/web/src/lib/model-pricing.ts
+++ b/web/src/lib/model-pricing.ts
@@ -1,5 +1,6 @@
 import { modelRequestOptions, resolveVideoOperation, type ModelRequirements } from "@/lib/model-selection";
 import { videoResolutionComparisonKey } from "@/lib/video-generation-options";
+import { buildImageResolutionOptions, imageResolutionOption } from "@/lib/image-resolution-tiers";
 import type { ModelRequestIntent } from "@/services/api/logical-models";
 import { modelOptionName, resolveModelChannel, type AiConfig, type ModelCapability } from "@/stores/use-config-store";
 
@@ -79,7 +80,7 @@ export function modelQuoteRequest(config: AiConfig, value: string, capability?:
     const input = requirements?.input;
     const intent: ModelRequestIntent = {
         capability,
-        operation: capability === "video" && input ? resolveVideoOperation(input, requirements?.videoOperation) : requirements?.videoOperation,
+        operation: capability === "image" ? imagePriceOperation(requirements) : capability === "video" && input ? resolveVideoOperation(input, requirements?.videoOperation) : requirements?.videoOperation,
         inputs: {
             image: (input?.imageCount || 0) + (input?.characterCount || 0),
             video: input?.videoCount || 0,
@@ -115,12 +116,27 @@ function priceSelectorForRequest(capability: ModelCapability | undefined, config
         if (seconds > 0) requested.videoSeconds = String(seconds);
     }
     if (capability === "image") {
-        if (config.quality && config.quality !== "auto") requested.quality = config.quality.toLowerCase();
-        if (config.size && config.size !== "auto") requested.size = config.size.toLowerCase();
+        requested.operation = imagePriceOperation(requirements);
+        const options = { ...modelRequestOptions(config, "image"), ...requirements?.options, ...(requirements?.imageSize ? { size: requirements.imageSize } : {}) };
+        const imageQuality = String(options.quality ?? "").trim().toLowerCase();
+        const imageSize = String(options.size ?? "").trim();
+        if ((imageQuality === "" || imageQuality === "auto" || imageQuality === "any") && imageSize) {
+            const resolution = imageResolutionOption(buildImageResolutionOptions([imageSize]), imageSize)?.tier;
+            if (resolution) requested.quality = resolution;
+        }
+        for (const key of ["quality", "size"] as const) {
+            const value = String(options[key] ?? "").trim().toLowerCase();
+            if (value && value !== "auto" && value !== "any" && !requested[key]) requested[key] = value;
+        }
     }
     return requested;
 }
 
+function imagePriceOperation(requirements?: ModelRequirements) {
+    const input = requirements?.input;
+    return (input?.imageCount || 0) + (input?.characterCount || 0) > 0 ? "image_to_image" : "text_to_image";
+}
+
 function priceSelectorForTier(tier: ModelPriceTier) {
     const selector = { ...(tier.selector || {}) };
     if (!Object.keys(selector).length) {
diff --git a/web/test/model-pricing.test.ts b/web/test/model-pricing.test.ts
index e20436421..139d9ffc6 100644
--- a/web/test/model-pricing.test.ts
+++ b/web/test/model-pricing.test.ts
@@ -3,6 +3,7 @@ import { describe, expect, test } from "bun:test";
 import { modelQuoteRequest, normalizeTierResolution, priceTierSummaryLabel, priceTiersForCurrentSelection, requestCreditCost } from "../src/lib/model-pricing";
 import type { ModelRequirements } from "../src/lib/model-selection";
 import { createModelChannel, defaultConfig, normalizeConfigSnapshot, resolveModelChannel, type AiConfig } from "../src/stores/use-config-store";
+import { resolvePricedImageQuality } from "../src/components/canvas/canvas-node-prompt-panel";
 
 function systemConfig(input: { capability?: "image" | "video"; logicalModelId?: string; tiers: Array<{ selector: Record<string, string>; billingMode: "fixed_request" | "per_second" | "token"; unitPriceMicrocredits: number }> }) {
     const capability = input.capability || "video";
@@ -58,6 +59,78 @@ const textVideoRequirements: ModelRequirements = {
 };
 
 describe("model request pricing", () => {
+    test("resolves automatic image quality to the first configured specification", () => {
+        const config = systemConfig({
+            capability: "image",
+            tiers: [
+                { selector: { quality: "1k" }, billingMode: "fixed_request", unitPriceMicrocredits: 1_000_000 },
+                { selector: { quality: "2k" }, billingMode: "fixed_request", unitPriceMicrocredits: 2_000_000 },
+                { selector: { quality: "4k" }, billingMode: "fixed_request", unitPriceMicrocredits: 4_000_000 },
+            ],
+        });
+        expect(resolvePricedImageQuality(config, config.model, "auto")).toBe("1k");
+        expect(resolvePricedImageQuality(config, config.model, "2k")).toBe("2k");
+    });
+
+    test("uses image dimensions to match resolution pricing when quality is automatic", () => {
+        const config = systemConfig({
+            capability: "image",
+            tiers: [
+                { selector: { quality: "1k" }, billingMode: "fixed_request", unitPriceMicrocredits: 2_000_000 },
+                { selector: { quality: "2k" }, billingMode: "fixed_request", unitPriceMicrocredits: 3_000_000 },
+                { selector: { quality: "4k" }, billingMode: "fixed_request", unitPriceMicrocredits: 5_000_000 },
+            ],
+        });
+        for (const [size, price] of [["1024x1024", 2], ["2048x2048", 3], ["2880x2880", 5]] as const) {
+            expect(requestCreditCost({ channelMode: "remote", modelCosts: resolveModelChannel(config, config.model).modelCosts, model: "image-model", capability: "image", config: { ...config, quality: "auto", size }, requirements: { capability: "image", options: { quality: "auto", size } }, count: 1 })).toBe(price);
+        }
+    });
+
+    test("matches each image resolution and generation operation for quotes and displayed costs", () => {
+        const qualities = ["1k", "2k", "4k"];
+        const config = systemConfig({
+            capability: "image",
+            logicalModelId: "logical-image",
+            tiers: ["text_to_image", "image_to_image"].flatMap((operation, operationIndex) =>
+                qualities.map((quality, index) => ({
+                    selector: { operation, quality },
+                    billingMode: "fixed_request" as const,
+                    unitPriceMicrocredits: (index + 1 + operationIndex * 3) * 1_000_000,
+                })),
+            ),
+        });
+        const channel = resolveModelChannel(config, config.model);
+        for (const imageCount of [0, 1]) {
+            for (const [index, quality] of qualities.entries()) {
+                const requirements: ModelRequirements = {
+                    capability: "image",
+                    input: { textCount: 1, imageCount, characterCount: 0, videoCount: 0, audioCount: 0 },
+                    options: { quality: quality.toUpperCase(), size: "16:9" },
+                };
+                expect(requestCreditCost({ channelMode: "remote", modelCosts: channel.modelCosts, model: "image-model", capability: "image", config, requirements, count: 1 })).toBe(index + 1 + imageCount * 3);
+                expect(modelQuoteRequest(config, config.model, "image", requirements)?.intent).toMatchObject({
+                    operation: imageCount ? "image_to_image" : "text_to_image",
+                    inputs: { image: imageCount },
+                    options: { quality: quality.toUpperCase() },
+                });
+            }
+        }
+        const characterRequirements: ModelRequirements = { capability: "image", input: { textCount: 1, imageCount: 0, characterCount: 1, videoCount: 0, audioCount: 0 }, options: { quality: "2k" } };
+        expect(modelQuoteRequest(config, config.model, "image", characterRequirements)?.intent).toMatchObject({ operation: "image_to_image", inputs: { image: 1 } });
+        expect(requestCreditCost({ channelMode: "remote", modelCosts: channel.modelCosts, model: "image-model", capability: "image", config, requirements: characterRequirements, count: 1 })).toBe(5);
+        expect(priceTiersForCurrentSelection(channel.modelCosts![0]!.logicalPriceTiers!, "image", { ...config, quality: "8k" })).toHaveLength(0);
+    });
+
+    test("prefers a matching image specification over the uniform fallback", () => {
+        const config = systemConfig({ capability: "image", tiers: [
+            { selector: {}, billingMode: "fixed_request", unitPriceMicrocredits: 1_000_000 },
+            { selector: { operation: "text_to_image", quality: "2k" }, billingMode: "fixed_request", unitPriceMicrocredits: 2_000_000 },
+        ] });
+        const tiers = resolveModelChannel(config, config.model).modelCosts![0]!.logicalPriceTiers!;
+        expect(priceTiersForCurrentSelection(tiers, "image", { ...config, quality: "2k" })[0]?.unitPriceMicrocredits).toBe(2_000_000);
+        expect(priceTiersForCurrentSelection(tiers, "image", { ...config, quality: "4k" })[0]?.unitPriceMicrocredits).toBe(1_000_000);
+    });
+
     test("preserves provider-specific resolution enums when matching price tiers", () => {
         expect(normalizeTierResolution("768P竖")).toBe("768p竖");
         expect(normalizeTierResolution("HD_Portrait")).toBe("hd_portrait");