feat(providers): add Gemini 4 Argon to Gemini and Vertex AI providers (#2043)

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

Author: zoomote[bot]

packages/types/src/__tests__/google-models.test.ts

@@ -1,6 +1,32 @@
 import { geminiModels } from "../providers/gemini.js"
 import { vertexModels } from "../providers/vertex.js"
 
+describe.each([
+	["Gemini API", geminiModels["gemini-4-argon"]],
+	["Vertex AI", vertexModels["gemini-4-argon"]],
+])("Gemini 4 Argon on %s", (_provider, model) => {
+	// https://cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-4-argon
+	it("exposes the documented limits, thinking levels, and promo pricing", () => {
+		expect(model.maxTokens).toBe(65_536)
+		expect(model.contextWindow).toBe(1_048_576)
+		expect(model.supportsImages).toBe(true)
+		// Implicit caching only; explicit context caching is unsupported.
+		expect(model.supportsPromptCache).toBe(true)
+		expect(model).not.toHaveProperty("cacheWritesPrice")
+		// LOW, MEDIUM (default), HIGH; thinking_budget and MINIMAL error out.
+		expect(model.supportsReasoningEffort).toEqual(["low", "medium", "high"])
+		expect(model.reasoningEffort).toBe("medium")
+		expect(model.supportsReasoningBudget).toBe(false)
+		// The backend ignores temperature/top_p/top_k.
+		expect(model.supportsTemperature).toBe(false)
+		// Promo pricing through Nov 30, 2026 (input $4 / output $20 / cache
+		// $0.20 per 1M tokens from Dec 1, 2026).
+		expect(model.inputPrice).toBe(2.0)
+		expect(model.outputPrice).toBe(10.0)
+		expect(model.cacheReadsPrice).toBe(0.1)
+	})
+})
+
 describe.each([
 	["Gemini API", geminiModels["gemini-3.7-flash"]],
 	["Vertex AI", vertexModels["gemini-3.7-flash"]],

packages/types/src/providers/gemini.ts

@@ -6,6 +6,26 @@ export type GeminiModelId = keyof typeof geminiModels
 export const geminiDefaultModelId: GeminiModelId = "gemini-3.1-pro-preview"
 
 export const geminiModels = {
+	// https://cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-4-argon
+	"gemini-4-argon": {
+		maxTokens: 65_536, // Single-request streaming cap; long decodes up to 1M via continuation tokens.
+		contextWindow: 1_048_576, // Shared across input, thinking, and output tokens.
+		supportsImages: true,
+		// Implicit caching only (automatic for prompt prefixes >= 6,144 tokens);
+		// explicit context caching is not supported on this model.
+		supportsPromptCache: true,
+		supportsReasoningEffort: ["low", "medium", "high"],
+		reasoningEffort: "medium",
+		// thinking_budget returns an HTTP 400 error on this model.
+		supportsReasoningBudget: false,
+		// The backend ignores temperature/top_p/top_k.
+		supportsTemperature: false,
+		// Promo pricing through Nov 30, 2026; from Dec 1, 2026: input $4.00,
+		// output $20.00, cached input $0.20 per 1M tokens.
+		inputPrice: 2.0,
+		outputPrice: 10.0, // Includes thinking tokens.
+		cacheReadsPrice: 0.1,
+	},
 	"gemini-3.7-flash": {
 		maxTokens: 65_536,
 		contextWindow: 1_048_576,

packages/types/src/providers/vertex.ts

@@ -6,6 +6,28 @@ export type VertexModelId = keyof typeof vertexModels
 export const vertexDefaultModelId: VertexModelId = "claude-sonnet-4-5@20250929"
 
 export const vertexModels = {
+	// https://cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-4-argon
+	// Available in the global, us, and eu locations.
+	"gemini-4-argon": {
+		maxTokens: 65_536, // Single-request streaming cap; long decodes up to 1M via continuation tokens.
+		contextWindow: 1_048_576, // Shared across input, thinking, and output tokens.
+		supportsImages: true,
+		// Implicit caching only (automatic for prompt prefixes >= 6,144 tokens);
+		// explicit context caching is not supported on this model.
+		supportsPromptCache: true,
+		supportsReasoningEffort: ["low", "medium", "high"],
+		reasoningEffort: "medium",
+		// thinking_budget returns an HTTP 400 error on this model.
+		supportsReasoningBudget: false,
+		// The backend ignores temperature/top_p/top_k.
+		supportsTemperature: false,
+		// Standard tier promo pricing (Global) through Nov 30, 2026; from
+		// Dec 1, 2026: input $4.00, output $20.00, cached input $0.20 per 1M
+		// tokens. Non-global locations are +10%.
+		inputPrice: 2.0,
+		outputPrice: 10.0, // Includes thinking tokens.
+		cacheReadsPrice: 0.1,
+	},
 	"gemini-3.7-flash": {
 		maxTokens: 65_536,
 		contextWindow: 1_048_576,