feat(minimax): add MiniMax M3 models (512K / 1M context) and fix double-think on Anthropic endpoint

packages/types/src/__tests__/provider-settings.test.ts

@@ -1,3 +1,4 @@
+import { describe, expect, it } from "vitest"
 import { getApiProtocol } from "../provider-settings.js"
 
 describe("getApiProtocol", () => {
@@ -88,6 +89,17 @@ describe("getApiProtocol", () => {
 		})
 	})
 
+	describe("MiniMax provider", () => {
+		it("should return 'anthropic' for MiniMax M3 and M2.x models", () => {
+			expect(getApiProtocol("minimax", "MiniMax-M3-1M")).toBe("anthropic")
+			expect(getApiProtocol("minimax", "MiniMax-M3-512k")).toBe("anthropic")
+			expect(getApiProtocol("minimax", "MiniMax-M2.7")).toBe("anthropic")
+			expect(getApiProtocol("minimax", "MiniMax-M2.5")).toBe("anthropic")
+			expect(getApiProtocol("minimax", "MiniMax-M2")).toBe("anthropic")
+			expect(getApiProtocol("minimax")).toBe("anthropic")
+		})
+	})
+
 	describe("Edge cases", () => {
 		it("should return 'openai' when provider is undefined", () => {
 			expect(getApiProtocol(undefined)).toBe("openai")

packages/types/src/providers/minimax.ts

@@ -5,7 +5,7 @@ import type { ModelInfo } from "../model.js"
 // https://platform.minimax.io/docs/guides/pricing-paygo
 // https://platform.minimax.io/docs/guides/pricing-tokenplan
 export type MinimaxModelId = keyof typeof minimaxModels
-export const minimaxDefaultModelId: MinimaxModelId = "MiniMax-M2.7"
+export const minimaxDefaultModelId: MinimaxModelId = "MiniMax-M3-1M"
 
 export const minimaxModels = {
 	"MiniMax-M2.5": {
@@ -68,6 +68,44 @@ export const minimaxModels = {
 		description:
 			"MiniMax M2.7 highspeed: same performance as M2.7 but with faster response (approximately 100 tps vs 60 tps). See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Requires TokenPlan High-Speed subscription for use with TokenPlan keys. Note: When using TokenPlan, usage is billed per request, not per token.",
 	},
+	"MiniMax-M3-512k": {
+		maxTokens: 65_536,
+		contextWindow: 524_288,
+		supportsImages: true,
+		supportsPromptCache: true,
+		includedTools: ["search_and_replace"],
+		excludedTools: ["apply_diff"],
+		preserveReasoning: true,
+		inputPrice: 0.3,
+		outputPrice: 1.2,
+		// M3 routes through the Anthropic Messages path on api.minimax.io/anthropic
+		// with client-side cache_control injection active, so cache_creation_input_tokens
+		// are reported and billed. Matches the MiniMax write price shared by the M2
+		// family (same vendor/pricing tier: $0.3 in / $1.2 out / $0.06 cache read).
+		cacheWritesPrice: 0.375,
+		cacheReadsPrice: 0.06,
+		description:
+			"MiniMax M3 (512K context) — the latest MiniMax model with multimodal input, stronger agentic reasoning and tool use, exposed via a 512K-token context window. See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Note: When using TokenPlan, usage is billed per request, not per token.",
+	},
+	"MiniMax-M3-1M": {
+		maxTokens: 131_072,
+		contextWindow: 1_048_576,
+		supportsImages: true,
+		supportsPromptCache: true,
+		includedTools: ["search_and_replace"],
+		excludedTools: ["apply_diff"],
+		preserveReasoning: true,
+		inputPrice: 0.3,
+		outputPrice: 1.2,
+		// M3 routes through the Anthropic Messages path on api.minimax.io/anthropic
+		// with client-side cache_control injection active, so cache_creation_input_tokens
+		// are reported and billed. Matches the MiniMax write price shared by the M2
+		// family (same vendor/pricing tier: $0.3 in / $1.2 out / $0.06 cache read).
+		cacheWritesPrice: 0.375,
+		cacheReadsPrice: 0.06,
+		description:
+			"MiniMax M3 (1M context) — the latest MiniMax model with multimodal input, stronger agentic reasoning and tool use, exposed via a 1M-token context window. See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Note: When using TokenPlan, usage is billed per request, not per token.",
+	},
 	"MiniMax-M2": {
 		maxTokens: 16_384,
 		contextWindow: 204_800,