feat(providers): add DeepSeek V4 Flash Vision Exp (#1488)

Co-authored-by: @taltas <6816042+taltas@users.noreply.github.com>

Author: zoomote[bot]

packages/types/src/providers/deepseek.ts

@@ -49,12 +49,14 @@ export const deepSeekModels = {
 		supportsReasoningEffort: ["disable", "low", "high", "max"], // Updated 2026-08-13
 		preserveReasoning: true,
 		reasoningEffort: "high",
+		supportsTemperature: true,
+		defaultTemperature: 1.0,
 		inputPrice: 0, // the inputs are priced as cache read/write, so `inputPrice` should be 0
 		// Static estimates use peak rates; off-peak rates are 50% lower.
 		outputPrice: 1.32,
 		cacheWritesPrice: 0.44,
 		cacheReadsPrice: 0.014,
-		description: `DeepSeek-V4-Flash-Vision-Exp is DeepSeek's experimental multimodal V4 Flash model with image understanding. It supports thinking and non-thinking modes, JSON output, tool calls, chat prefix completion (beta), and image inputs.`,
+		description: `DeepSeek-V4-Flash-Vision-Exp is DeepSeek's experimental multimodal V4 Flash model with image understanding. It supports thinking and non-thinking modes, JSON output, tool calls, chat prefix completion (beta), and image input through Chat Completions, Responses, and Anthropic-compatible APIs.`,
 	},
 } as const satisfies Record<string, ModelInfo>
 

packages/types/src/providers/fireworks.ts

@@ -17,6 +17,7 @@ export type FireworksModelId =
 	| "accounts/fireworks/models/deepseek-v3p2"
 	| "accounts/fireworks/models/deepseek-v4-pro"
 	| "accounts/fireworks/models/deepseek-v4-pro-0813"
+	| "accounts/fireworks/models/deepseek-v4-flash-vision-exp"
 	| "accounts/fireworks/models/glm-4p5"
 	| "accounts/fireworks/models/glm-4p5-air"
 	| "accounts/fireworks/models/glm-4p6"
@@ -280,6 +281,19 @@ export const fireworksModels = {
 		description:
 			"DeepSeek V4 Pro 0813 is DeepSeek's production checkpoint for advanced reasoning, coding, and long-context agentic workloads.",
 	},
+	"accounts/fireworks/models/deepseek-v4-flash-vision-exp": {
+		displayName: "DeepSeek V4 Flash Vision Exp",
+		maxTokens: 384_000,
+		contextWindow: 1_048_576,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		inputPrice: 0.22,
+		outputPrice: 0.66,
+		cacheReadsPrice: 0.007,
+		description:
+			"DeepSeek V4 Flash Vision Exp is an experimental multimodal model with text and image input, function calling, and long-context support.",
+	},
 	"accounts/fireworks/models/kimi-k2p7-code": {
 		maxTokens: 16384,
 		contextWindow: 262144,

src/api/index.ts

@@ -236,7 +236,7 @@ export function buildApiHandler(configuration: ProviderSettings): ApiHandler {
 		case providerIdentifiers.poe:
 			return new PoeHandler(options)
 		case providerIdentifiers.geminiCli:
-			// Intentionally falls through to the Anthropic handler pending a dedicated Gemini CLI handler implementation.
+		// Intentionally falls through to the Anthropic handler pending a dedicated Gemini CLI handler implementation.
 		default:
 			return new AnthropicHandler(options)
 	}

src/api/providers/__tests__/deepseek.spec.ts

@@ -249,16 +249,23 @@ describe("DeepSeekHandler", () => {
 			expect((model.info as ModelInfo).reasoningEffort).toBe("high")
 		})
 
-		it("should return vision model info for deepseek-v4-flash-vision-exp", () => {
-			const handlerWithVision = new DeepSeekHandler({
+		it("should return vision metadata for deepseek-v4-flash-vision-exp", () => {
+			const visionHandler = new DeepSeekHandler({
 				...mockOptions,
 				apiModelId: "deepseek-v4-flash-vision-exp",
 			})
-			const model = handlerWithVision.getModel()
+			const model = visionHandler.getModel()
+
 			expect(model.id).toBe("deepseek-v4-flash-vision-exp")
-			expect(model.info.supportsImages).toBe(true)
-			expect(model.info.supportsPromptCache).toBe(true)
-			expect((model.info as ModelInfo).preserveReasoning).toBe(true)
+			expect(model.info).toMatchObject({
+				maxTokens: 384_000,
+				contextWindow: 1_000_000,
+				supportsImages: true,
+				supportsPromptCache: true,
+				preserveReasoning: true,
+				reasoningEffort: "high",
+				defaultTemperature: 1.0,
+			})
 		})
 
 		it("should return provided model ID with default model info if model does not exist", () => {
@@ -336,6 +343,43 @@ describe("DeepSeekHandler", () => {
 			expect(textChunks[0].text).toBe("Test response")
 		})
 
+		it("should send images and V4 thinking controls to deepseek-v4-flash-vision-exp", async () => {
+			const visionHandler = new DeepSeekHandler({
+				...mockOptions,
+				apiModelId: "deepseek-v4-flash-vision-exp",
+			})
+			const visionMessages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "user",
+					content: [
+						{ type: "text", text: "Describe this image." },
+						{
+							type: "image",
+							source: { type: "base64", media_type: "image/png", data: "image-data" },
+						},
+					],
+				},
+			]
+
+			await collectStream(visionHandler.createMessage(systemPrompt, visionMessages))
+
+			const callArgs = mockCreate.mock.calls[0][0]
+			expect(callArgs).toMatchObject({
+				model: "deepseek-v4-flash-vision-exp",
+				thinking: { type: "enabled" },
+				reasoning_effort: "high",
+				max_completion_tokens: 200_000,
+			})
+			expect(callArgs.temperature).toBeUndefined()
+			expect(callArgs.messages).toContainEqual({
+				role: "user",
+				content: expect.arrayContaining([
+					{ type: "text", text: expect.stringContaining("Describe this image.") },
+					{ type: "image_url", image_url: { url: "data:image/png;base64,image-data" } },
+				]),
+			})
+		})
+
 		it("should include usage information", async () => {
 			const chunks: any[] = await collectStream(handler.createMessage(systemPrompt, messages))
 
@@ -424,42 +468,6 @@ describe("DeepSeekHandler", () => {
 			const reasoningChunks = chunks.filter((chunk) => chunk.type === "reasoning")
 			expect(reasoningChunks).toEqual([{ type: "reasoning", text: "primary thought" }])
 		})
-
-		it("enables thinking and forwards image_url for the vision model", async () => {
-			const visionHandler = new DeepSeekHandler({
-				...mockOptions,
-				apiModelId: "deepseek-v4-flash-vision-exp",
-			})
-			const withImage: Anthropic.Messages.MessageParam[] = [
-				{
-					role: "user",
-					content: [
-						{ type: "text", text: "Describe:" },
-						{ type: "image", source: { type: "base64", media_type: "image/png", data: "abc" } },
-					],
-				},
-			]
-
-			await collectStream(visionHandler.createMessage(systemPrompt, withImage))
-
-			const callArgs = mockCreate.mock.calls[0][0]
-			expect(callArgs.thinking).toEqual({ type: "enabled" })
-			expect(callArgs.messages).toEqual(
-				expect.arrayContaining([
-					expect.objectContaining({
-						role: "user",
-						content: expect.arrayContaining([
-							expect.objectContaining({
-								type: "image_url",
-								image_url: expect.objectContaining({
-									url: "data:image/png;base64,abc",
-								}),
-							}),
-						]),
-					}),
-				]),
-			)
-		})
 	})
 
 	describe("processUsageMetrics", () => {

src/api/providers/__tests__/fireworks.spec.ts

@@ -127,6 +127,13 @@ describe("FireworksHandler", () => {
 			outputPrice: 3.96,
 			cacheReadsPrice: 0.044,
 		},
+		{
+			modelId: "accounts/fireworks/models/deepseek-v4-flash-vision-exp" as const,
+			contextWindow: 1_048_576,
+			inputPrice: 0.22,
+			outputPrice: 0.66,
+			cacheReadsPrice: 0.007,
+		},
 	])(
 		"should expose newly added model $modelId",
 		({ modelId, contextWindow, inputPrice, outputPrice, cacheReadsPrice }) => {
@@ -147,6 +154,16 @@ describe("FireworksHandler", () => {
 		},
 	)
 
+	it("should expose vision support for DeepSeek V4 Flash Vision Exp", () => {
+		const model = fireworksModels["accounts/fireworks/models/deepseek-v4-flash-vision-exp"]
+
+		expect(model).toMatchObject({
+			supportsImages: true,
+			supportsPromptCache: true,
+			supportsMaxTokens: true,
+		})
+	})
+
 	it("should return Kimi K2 Instruct model with correct configuration", () => {
 		const testModelId: FireworksModelId = "accounts/fireworks/models/kimi-k2-instruct"
 		const handlerWithModel = new FireworksHandler({

src/api/providers/fetchers/__tests__/deepseek.spec.ts

@@ -31,6 +31,7 @@ describe("getDeepSeekModels", () => {
 		expect(globalThis.fetch).toHaveBeenCalledWith("http://127.0.0.1:43123/models", expect.any(Object))
 		expect(models["deepseek-v4-flash"]).toEqual(deepSeekModels["deepseek-v4-flash"])
 		expect(models["deepseek-v4-pro"]).toEqual(deepSeekModels["deepseek-v4-pro"])
+		expect(models["deepseek-v4-flash-vision-exp"]).toEqual(deepSeekModels["deepseek-v4-flash-vision-exp"])
 	})
 
 	it("throws for 404 responses when fallback flag is not enabled", async () => {

webview-ui/src/components/ui/hooks/__tests__/useSelectedModel.spec.ts

@@ -220,6 +220,21 @@ describe("useSelectedModel", () => {
 		},
 	)
 
+	it("selects static vision metadata when the DeepSeek catalog is unavailable", () => {
+		const modelId = "deepseek-v4-flash-vision-exp"
+		mockUseRouterModels.mockReturnValue(createRouterModelsResult({ [providerIdentifiers.deepseek]: null }))
+		mockUseOpenRouterModelProviders.mockReturnValue(createOpenRouterModelProvidersResult({}))
+
+		const { result } = renderHook(
+			() => useSelectedModel({ apiProvider: providerIdentifiers.deepseek, apiModelId: modelId }),
+			{ wrapper: createWrapper() },
+		)
+
+		expect(result.current.id).toBe(modelId)
+		expect(result.current.info).toEqual(deepSeekModels[modelId])
+		expect(result.current.info?.supportsImages).toBe(true)
+	})
+
 	it.each([providerIdentifiers.deepseek, providerIdentifiers.moonshot])(
 		"falls back to static data when the %s router catalog is null",
 		(provider) => {