diff --git a/packages/types/src/__tests__/opencode-go.test.ts b/packages/types/src/__tests__/opencode-go.test.ts
index c65de6165c..ecba7d5324 100644
--- a/packages/types/src/__tests__/opencode-go.test.ts
+++ b/packages/types/src/__tests__/opencode-go.test.ts
@@ -13,6 +13,7 @@ import {
 describe("opencode-go registry", () => {
 	const anthropicFormatModels = [
 		"qwen3.8-max",
+		"qwen3.8-flash",
 		"qwen3.7-max",
 		"qwen3.7-plus",
 		"qwen3.6-plus",
@@ -33,6 +34,13 @@ describe("opencode-go registry", () => {
 		"deepseek-v4-pro",
 		"deepseek-v4-flash",
 	]
+	const responsesFormatModels = [
+		"gpt-5.6-luna",
+		"grok-4.5",
+		"grok-4.6",
+		"muse-spark-1.3-contributor",
+		"muse-spark-1.2-contributor",
+	]
 
 	describe("isOpencodeGoAnthropicFormatModel", () => {
 		it("classifies Qwen and MiniMax models as Anthropic-format", () => {
@@ -133,11 +141,13 @@ describe("opencode-go registry", () => {
 
 	describe("OPENCODE_GO_RESPONSES_FORMAT_MODELS", () => {
 		it("contains exactly the Responses-only models", () => {
-			expect([...OPENCODE_GO_RESPONSES_FORMAT_MODELS].sort()).toEqual(["gpt-5.6-luna"])
+			expect([...OPENCODE_GO_RESPONSES_FORMAT_MODELS].sort()).toEqual([...responsesFormatModels].sort())
 		})
 
-		it("classifies gpt-5.6-luna as Responses-format", () => {
-			expect(isOpencodeGoResponsesFormatModel("gpt-5.6-luna")).toBe(true)
+		it("classifies every Responses-only model as Responses-format", () => {
+			for (const id of responsesFormatModels) {
+				expect(isOpencodeGoResponsesFormatModel(id)).toBe(true)
+			}
 		})
 
 		it("classifies Anthropic-format and OpenAI-compatible models as non-Responses-format", () => {
@@ -156,6 +166,7 @@ describe("opencode-go registry", () => {
 
 		it("curates gpt-5.6-luna with its Go Responses capabilities", () => {
 			expect(getOpencodeGoModelInfo("gpt-5.6-luna")).toMatchObject({
+				supportsMaxTokens: true,
 				maxTokens: 128_000,
 				contextWindow: 1_050_000,
 				supportsImages: true,
@@ -184,6 +195,195 @@ describe("opencode-go registry", () => {
 	})
 
 	describe("opencodeGoModels registry invariants", () => {
+		it.each([
+			{
+				id: "glm-5.3-flash",
+				expected: {
+					maxTokens: 131_072,
+					contextWindow: 1_000_000,
+					supportsImages: true,
+					supportsPromptCache: true,
+					supportsMaxTokens: true,
+					supportsReasoningEffort: ["low", "high", "max"],
+					inputPrice: 0.075,
+					outputPrice: 0.25,
+					cacheReadsPrice: 0.015,
+				},
+			},
+			{
+				id: "kimi-k2.7-code",
+				expected: {
+					maxTokens: 262_144,
+					contextWindow: 262_144,
+					supportsImages: true,
+					supportsPromptCache: true,
+					supportsMaxTokens: true,
+					inputPrice: 0.95,
+					outputPrice: 4,
+					cacheReadsPrice: 0.19,
+				},
+			},
+			{
+				id: "longcat-2.0",
+				expected: {
+					maxTokens: 131_072,
+					contextWindow: 1_000_000,
+					supportsImages: false,
+					supportsReasoningBinary: true,
+					inputPrice: 0.3,
+					outputPrice: 1.2,
+					cacheReadsPrice: 0.006,
+				},
+			},
+			{
+				id: "mimo-v2-pro",
+				expected: {
+					maxTokens: 128_000,
+					contextWindow: 1_048_576,
+					supportsImages: false,
+					supportsPromptCache: false,
+					inputPrice: 1,
+					outputPrice: 3,
+					cacheReadsPrice: 0.2,
+					longContextPricing: {
+						thresholdTokens: 256_000,
+						inputPriceMultiplier: 2,
+						outputPriceMultiplier: 2,
+						cacheReadsPriceMultiplier: 2,
+					},
+				},
+			},
+			{
+				id: "mimo-v2-omni",
+				expected: {
+					maxTokens: 128_000,
+					contextWindow: 262_144,
+					supportsImages: true,
+					supportsPromptCache: false,
+					inputPrice: 0.4,
+					outputPrice: 2,
+					cacheReadsPrice: 0.08,
+				},
+			},
+			{
+				id: "qwen3.5-plus",
+				expected: {
+					maxTokens: 65_536,
+					contextWindow: 262_144,
+					supportsImages: true,
+					supportsReasoningBudget: true,
+					supportsReasoningBinary: true,
+					inputPrice: 0.2,
+					outputPrice: 1.2,
+					cacheReadsPrice: 0.02,
+					cacheWritesPrice: 0.25,
+				},
+			},
+			{
+				id: "qwen3.8-flash",
+				expected: {
+					maxTokens: 131_072,
+					contextWindow: 1_000_000,
+					supportsImages: true,
+					supportsMaxTokens: true,
+					supportsReasoningBudget: true,
+					supportsReasoningBinary: true,
+					inputPrice: 0.15,
+					outputPrice: 0.47,
+					cacheReadsPrice: 0.016,
+					cacheWritesPrice: 0.2,
+				},
+			},
+			{
+				id: "deepseek-v4-flash-vision-exp",
+				expected: {
+					maxTokens: 384_000,
+					contextWindow: 1_000_000,
+					supportsImages: true,
+					supportsReasoningEffort: ["disable", "low", "high", "max"],
+					inputPrice: 0.22,
+					outputPrice: 0.66,
+					cacheReadsPrice: 0.007,
+				},
+			},
+			{
+				id: "hy4-preview",
+				expected: {
+					maxTokens: 64_000,
+					contextWindow: 1_024_000,
+					supportsImages: false,
+					supportsReasoningEffort: ["disable", "high"],
+					inputPrice: 0.834,
+					outputPrice: 2.501,
+					cacheReadsPrice: 0.042,
+				},
+			},
+			...(["hy3", "hy3-preview"] as const).map((id) => ({
+				id,
+				expected: {
+					maxTokens: 64_000,
+					contextWindow: 256_000,
+					supportsImages: false,
+					supportsReasoningEffort: ["disable", "low", "high"],
+					inputPrice: 0.0175,
+					outputPrice: 0.0725,
+					cacheReadsPrice: 0.004375,
+				},
+			})),
+			{
+				id: "grok-4.5",
+				expected: {
+					maxTokens: 500_000,
+					contextWindow: 500_000,
+					supportsImages: true,
+					supportsReasoningEffort: ["low", "medium", "high"],
+					inputPrice: 2,
+					outputPrice: 6,
+					cacheReadsPrice: 0.3,
+					longContextPricing: {
+						thresholdTokens: 200_000,
+						inputPriceMultiplier: 2,
+						outputPriceMultiplier: 2,
+						cacheReadsPriceMultiplier: 2,
+					},
+				},
+			},
+			{
+				id: "grok-4.6",
+				expected: {
+					maxTokens: 500_000,
+					contextWindow: 500_000,
+					supportsImages: true,
+					supportsReasoningEffort: ["low", "medium", "high", "xhigh"],
+					inputPrice: 2,
+					outputPrice: 6,
+					cacheReadsPrice: 0.5,
+					longContextPricing: {
+						thresholdTokens: 200_000,
+						inputPriceMultiplier: 2,
+						outputPriceMultiplier: 2,
+						cacheReadsPriceMultiplier: 2,
+					},
+				},
+			},
+			...(["muse-spark-1.3-contributor", "muse-spark-1.2-contributor"] as const).map((id) => ({
+				id,
+				expected: {
+					maxTokens: 131_072,
+					contextWindow: 1_048_576,
+					supportsImages: true,
+					supportsPromptCache: true,
+					supportsMaxTokens: true,
+					supportsReasoningEffort: ["minimal", "low", "medium", "high", "xhigh"],
+					inputPrice: 0.1,
+					outputPrice: 0.2,
+					cacheReadsPrice: 0.002,
+				},
+			})),
+		])("keeps independently asserted metadata for $id", ({ id, expected }) => {
+			expect(getOpencodeGoModelInfo(id)).toMatchObject(expected)
+		})
+
 		it("every entry has a positive maxTokens and contextWindow", () => {
 			for (const [id, info] of Object.entries(opencodeGoModels)) {
 				expect(info.maxTokens).toBeGreaterThan(0)
diff --git a/packages/types/src/providers/opencode-go.ts b/packages/types/src/providers/opencode-go.ts
index 10d6bcbc50..eb32dbcac5 100644
--- a/packages/types/src/providers/opencode-go.ts
+++ b/packages/types/src/providers/opencode-go.ts
@@ -3,9 +3,9 @@ import type { ModelInfo } from "../model.js"
 // Opencode "Go" plan — OpenAI-compatible gateway.
 // https://opencode.ai/docs/go/ · base URL: https://opencode.ai/zen/go/v1
 //
-// The full model list (and metadata) is fetched dynamically from
-// `https://opencode.ai/zen/go/v1/models`, so models can be switched on the fly.
-// The values below are only a fallback used before the live list resolves.
+// Model IDs are fetched dynamically from `https://opencode.ai/zen/go/v1/models`
+// so models can be switched on the fly. The endpoint currently omits metadata,
+// so Zoo Code supplies limits and capabilities from the registries below.
 export const opencodeGoDefaultModelId = "glm-5.2"
 
 export const opencodeGoDefaultModelInfo: ModelInfo = {
@@ -16,7 +16,7 @@ export const opencodeGoDefaultModelInfo: ModelInfo = {
 	// Pricing is intentionally omitted: ModelInfoView renders a `0` field as "$0.00 / 1M tokens"
 	// (implying the service is free), so we leave it unknown — consistent with the dynamically
 	// fetched models, which also leave price fields absent. See PR #319 review.
-	description: "Opencode Go plan model. Available models and metadata are resolved dynamically from /v1/models.",
+	description: "Opencode Go plan model. Model IDs are fetched dynamically; metadata uses Zoo Code's model registry.",
 }
 
 export const OPENCODE_GO_DEFAULT_TEMPERATURE = 0
@@ -102,6 +102,21 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 		description:
 			"GLM-5.3 is Zhipu's flagship coding and agent model with a 1M context window, 128k max output, and always-on reasoning with configurable effort (Low/High/Max). Available via the Opencode Go plan.",
 	},
+	"glm-5.3-flash": {
+		maxTokens: 131_072,
+		contextWindow: 1_000_000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["low", "high", "max"],
+		reasoningEffort: "max",
+		preserveReasoning: true,
+		inputPrice: 0.075,
+		outputPrice: 0.25,
+		cacheReadsPrice: 0.015,
+		description:
+			"GLM-5.3 Flash is Zhipu's fast multimodal coding and agent model with a 1M context window and configurable reasoning effort. Available via the Opencode Go plan.",
+	},
 	"glm-5.2": {
 		maxTokens: 131_072,
 		contextWindow: 1_000_000,
@@ -163,6 +178,36 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 		description:
 			"Kimi K2.6 is Moonshot AI's native multimodal agentic MoE model with a 256k context window, built for long-horizon coding and tool use. Available via the Opencode Go plan.",
 	},
+	"kimi-k2.7-code": {
+		maxTokens: 262_144,
+		contextWindow: 262_144,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		preserveReasoning: true,
+		defaultTemperature: 1.0,
+		inputPrice: 0.95,
+		outputPrice: 4.0,
+		cacheReadsPrice: 0.19,
+		description:
+			"Kimi K2.7 Code is Moonshot AI's coding model for long-context programming tasks, with multimodal input and a 256k context window. Available via the Opencode Go plan.",
+	},
+
+	// --- Meituan LongCat ---
+	"longcat-2.0": {
+		maxTokens: 131_072,
+		contextWindow: 1_000_000,
+		supportsImages: false,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningBinary: true,
+		preserveReasoning: true,
+		inputPrice: 0.3,
+		outputPrice: 1.2,
+		cacheReadsPrice: 0.006,
+		description:
+			"LongCat 2.0 is Meituan's long-context reasoning and coding model with a 1M context window. Available via the Opencode Go plan.",
+	},
 
 	// --- Xiaomi MiMo ---
 	"mimo-v2.5": {
@@ -201,6 +246,36 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 		description:
 			"MiMo V2.5 Pro - Xiaomi's flagship reasoning model with 1M context, deep thinking, and tool calling. Available via the Opencode Go plan.",
 	},
+	"mimo-v2-pro": {
+		maxTokens: 128_000,
+		contextWindow: 1_048_576,
+		supportsImages: false,
+		supportsPromptCache: false,
+		preserveReasoning: true,
+		inputPrice: 1.0,
+		outputPrice: 3.0,
+		cacheReadsPrice: 0.2,
+		longContextPricing: {
+			thresholdTokens: 256_000,
+			inputPriceMultiplier: 2,
+			outputPriceMultiplier: 2,
+			cacheReadsPriceMultiplier: 2,
+		},
+		description:
+			"MiMo V2 Pro is Xiaomi's text reasoning and coding model with a 1M context window. Available via the Opencode Go plan.",
+	},
+	"mimo-v2-omni": {
+		maxTokens: 128_000,
+		contextWindow: 262_144,
+		supportsImages: true,
+		supportsPromptCache: false,
+		preserveReasoning: true,
+		inputPrice: 0.4,
+		outputPrice: 2.0,
+		cacheReadsPrice: 0.08,
+		description:
+			"MiMo V2 Omni is Xiaomi's multimodal reasoning model with a 256k context window. Available via the Opencode Go plan.",
+	},
 
 	// --- MiniMax ---
 	"minimax-m2.5": {
@@ -255,6 +330,21 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 	},
 
 	// --- Alibaba Qwen ---
+	"qwen3.5-plus": {
+		maxTokens: 65_536,
+		contextWindow: 262_144,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsReasoningBudget: true,
+		supportsReasoningBinary: true,
+		preserveReasoning: true,
+		inputPrice: 0.2,
+		outputPrice: 1.2,
+		cacheReadsPrice: 0.02,
+		cacheWritesPrice: 0.25,
+		description:
+			"Qwen3.5 Plus is Alibaba's multimodal reasoning model with a 256k context window. Available via the Opencode Go plan.",
+	},
 	"qwen3.6-plus": {
 		maxTokens: 65_536,
 		contextWindow: 1_000_000,
@@ -321,6 +411,22 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 		description:
 			"Qwen3.8 Max - Alibaba's flagship multimodal reasoning model with a 1M context window, 128k max output, and long-horizon coding and agentic capabilities. Available via the Opencode Go plan.",
 	},
+	"qwen3.8-flash": {
+		maxTokens: 131_072,
+		contextWindow: 1_000_000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningBudget: true,
+		supportsReasoningBinary: true,
+		preserveReasoning: true,
+		inputPrice: 0.15,
+		outputPrice: 0.47,
+		cacheReadsPrice: 0.016,
+		cacheWritesPrice: 0.2,
+		description:
+			"Qwen3.8 Flash is Alibaba's fast multimodal reasoning model with a 1M context window. Available via the Opencode Go plan.",
+	},
 
 	// --- DeepSeek ---
 	"deepseek-v4-pro": {
@@ -359,6 +465,70 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 		description:
 			"DeepSeek-V4-Flash is DeepSeek's fast, cost-efficient V4 model supporting thinking and non-thinking modes. Available via the Opencode Go plan.",
 	},
+	"deepseek-v4-flash-vision-exp": {
+		maxTokens: 384_000,
+		contextWindow: 1_000_000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["disable", "low", "high", "max"],
+		reasoningEffort: "high",
+		preserveReasoning: true,
+		inputPrice: 0.22,
+		outputPrice: 0.66,
+		cacheReadsPrice: 0.007,
+		description:
+			"DeepSeek V4 Flash Vision Experimental is a fast multimodal reasoning model with a 1M context window. Available via the Opencode Go plan.",
+	},
+
+	// --- Tencent Hunyuan ---
+	"hy4-preview": {
+		maxTokens: 64_000,
+		contextWindow: 1_024_000,
+		supportsImages: false,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["disable", "high"],
+		reasoningEffort: "high",
+		preserveReasoning: true,
+		inputPrice: 0.834,
+		outputPrice: 2.501,
+		cacheReadsPrice: 0.042,
+		description:
+			"Hunyuan 4 Preview is Tencent's long-context reasoning and coding model with a 1M context window. Available via the Opencode Go plan.",
+	},
+	hy3: {
+		maxTokens: 64_000,
+		contextWindow: 256_000,
+		supportsImages: false,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["disable", "low", "high"],
+		reasoningEffort: "high",
+		preserveReasoning: true,
+		inputPrice: 0.0175,
+		outputPrice: 0.0725,
+		cacheReadsPrice: 0.004375,
+		description:
+			"Hunyuan 3 is Tencent's reasoning and coding model with a 256k context window. Available via the Opencode Go plan.",
+	},
+	"hy3-preview": {
+		// The live endpoint still lists this historical alias; it shares HY3's limits and pricing.
+		maxTokens: 64_000,
+		contextWindow: 256_000,
+		supportsImages: false,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["disable", "low", "high"],
+		reasoningEffort: "high",
+		preserveReasoning: true,
+		inputPrice: 0.0175,
+		outputPrice: 0.0725,
+		cacheReadsPrice: 0.004375,
+		description:
+			"Hunyuan 3 Preview is Tencent's preview reasoning and coding model with a 256k context window. Available via the Opencode Go plan.",
+	},
+
 	// --- OpenAI Responses ---
 	// Luna is curated here because the Go gateway's model catalogue does not
 	// currently provide its capability metadata. These values intentionally
@@ -368,6 +538,7 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 		contextWindow: 1_050_000,
 		supportsImages: true,
 		supportsPromptCache: true,
+		supportsMaxTokens: true,
 		supportsReasoningEffort: ["none", "low", "medium", "high", "xhigh", "max"],
 		reasoningEffort: "medium",
 		inputPrice: 0.2,
@@ -381,7 +552,76 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 			cacheWritesPriceMultiplier: 2,
 			cacheReadsPriceMultiplier: 2,
 		},
-		description: "GPT-5.6 Luna via the OpenCode Go Responses API.",
+		description:
+			"GPT-5.6 Luna is OpenAI's fast reasoning model with a 1M context window. Available via the Opencode Go plan.",
+	},
+	"grok-4.5": {
+		maxTokens: 500_000,
+		contextWindow: 500_000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["low", "medium", "high"],
+		reasoningEffort: "high",
+		inputPrice: 2.0,
+		outputPrice: 6.0,
+		cacheReadsPrice: 0.3,
+		longContextPricing: {
+			thresholdTokens: 200_000,
+			inputPriceMultiplier: 2,
+			outputPriceMultiplier: 2,
+			cacheReadsPriceMultiplier: 2,
+		},
+		description:
+			"Grok 4.5 is xAI's multimodal reasoning and agent model with a 500k context window. Available via the Opencode Go plan.",
+	},
+	"grok-4.6": {
+		maxTokens: 500_000,
+		contextWindow: 500_000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["low", "medium", "high", "xhigh"],
+		reasoningEffort: "high",
+		inputPrice: 2.0,
+		outputPrice: 6.0,
+		cacheReadsPrice: 0.5,
+		longContextPricing: {
+			thresholdTokens: 200_000,
+			inputPriceMultiplier: 2,
+			outputPriceMultiplier: 2,
+			cacheReadsPriceMultiplier: 2,
+		},
+		description:
+			"Grok 4.6 is xAI's multimodal reasoning and agent model with a 500k context window. Available via the Opencode Go plan.",
+	},
+	"muse-spark-1.3-contributor": {
+		maxTokens: 131_072,
+		contextWindow: 1_048_576,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["minimal", "low", "medium", "high", "xhigh"],
+		reasoningEffort: "medium",
+		inputPrice: 0.1,
+		outputPrice: 0.2,
+		cacheReadsPrice: 0.002,
+		description:
+			"Muse Spark 1.3 Contributor is Meta's multimodal reasoning model for coding and agentic workflows. Available via the Opencode Go plan.",
+	},
+	"muse-spark-1.2-contributor": {
+		maxTokens: 131_072,
+		contextWindow: 1_048_576,
+		supportsImages: true,
+		supportsPromptCache: true,
+		supportsMaxTokens: true,
+		supportsReasoningEffort: ["minimal", "low", "medium", "high", "xhigh"],
+		reasoningEffort: "medium",
+		inputPrice: 0.1,
+		outputPrice: 0.2,
+		cacheReadsPrice: 0.002,
+		description:
+			"Muse Spark 1.2 Contributor is Meta's multimodal coding model with a 1M context window. Available via the Opencode Go plan.",
 	},
 }
 
@@ -404,6 +644,7 @@ export const opencodeGoModels: Record<string, ModelInfo> = {
 export const OPENCODE_GO_ANTHROPIC_FORMAT_MODELS = new Set<string>([
 	// --- Alibaba Qwen ---
 	"qwen3.8-max",
+	"qwen3.8-flash",
 	"qwen3.7-max",
 	"qwen3.7-plus",
 	"qwen3.6-plus",
@@ -418,10 +659,9 @@ export const OPENCODE_GO_ANTHROPIC_FORMAT_MODELS = new Set<string>([
  * (`/v1/responses`), not the OpenAI-compatible Chat Completions endpoint
  * (`/v1/chat/completions`).
  *
- * The Go gateway maps every model to exactly one wire format. Some models
- * (currently only `gpt-5.6-luna`) are Responses-only and are also explicitly
- * curated in `opencodeGoModels`: the gateway's
- * `/v1/chat/completions` adapter for them fails with an opaque HTTP 500
+ * The Go gateway maps every model to exactly one wire format. Responses-only
+ * models are explicitly curated in `opencodeGoModels`: the gateway's
+ * `/v1/chat/completions` adapter for these models can fail with an opaque HTTP 500
  * (`{"type":"error","error":{"type":"error","message":"Internal server error"}}`),
  * while `/v1/responses` succeeds (Zoo-Code-Org/Zoo-Code#1431).
  *
@@ -433,6 +673,12 @@ export const OPENCODE_GO_ANTHROPIC_FORMAT_MODELS = new Set<string>([
 export const OPENCODE_GO_RESPONSES_FORMAT_MODELS = new Set<string>([
 	// --- OpenAI ---
 	"gpt-5.6-luna",
+	// --- xAI ---
+	"grok-4.5",
+	"grok-4.6",
+	// --- Meta ---
+	"muse-spark-1.3-contributor",
+	"muse-spark-1.2-contributor",
 ])
 
 /**
diff --git a/src/api/providers/__tests__/opencode-go.spec.ts b/src/api/providers/__tests__/opencode-go.spec.ts
index 7415ea73b1..964299a154 100644
--- a/src/api/providers/__tests__/opencode-go.spec.ts
+++ b/src/api/providers/__tests__/opencode-go.spec.ts
@@ -1362,8 +1362,12 @@ describe("OpencodeGoHandler", () => {
 			}).rejects.toThrow("Opencode Go completion error: internal server error")
 		})
 
-		it("classifies gpt-5.6-luna as Responses-format and other models as not", () => {
+		it("classifies documented Responses models as Responses-format and other models as not", () => {
 			expect(isOpencodeGoResponsesFormatModel("gpt-5.6-luna")).toBe(true)
+			expect(isOpencodeGoResponsesFormatModel("grok-4.5")).toBe(true)
+			expect(isOpencodeGoResponsesFormatModel("grok-4.6")).toBe(true)
+			expect(isOpencodeGoResponsesFormatModel("muse-spark-1.3-contributor")).toBe(true)
+			expect(isOpencodeGoResponsesFormatModel("muse-spark-1.2-contributor")).toBe(true)
 			expect(isOpencodeGoResponsesFormatModel("glm-5.3")).toBe(false)
 			expect(isOpencodeGoResponsesFormatModel("qwen3.7-max")).toBe(false)
 			expect(isOpencodeGoResponsesFormatModel("some-unknown-model")).toBe(false)
@@ -1372,6 +1376,7 @@ describe("OpencodeGoHandler", () => {
 
 	describe("isOpencodeGoAnthropicFormatModel", () => {
 		it("classifies Qwen and MiniMax Go models as Anthropic-format", () => {
+			expect(isOpencodeGoAnthropicFormatModel("qwen3.8-flash")).toBe(true)
 			expect(isOpencodeGoAnthropicFormatModel("qwen3.7-max")).toBe(true)
 			expect(isOpencodeGoAnthropicFormatModel("qwen3.7-plus")).toBe(true)
 			expect(isOpencodeGoAnthropicFormatModel("qwen3.6-plus")).toBe(true)
@@ -1385,6 +1390,7 @@ describe("OpencodeGoHandler", () => {
 			expect(isOpencodeGoAnthropicFormatModel("kimi-k2.6")).toBe(false)
 			expect(isOpencodeGoAnthropicFormatModel("deepseek-v4-pro")).toBe(false)
 			expect(isOpencodeGoAnthropicFormatModel("mimo-v2.5")).toBe(false)
+			expect(isOpencodeGoAnthropicFormatModel("qwen3.5-plus")).toBe(false)
 		})
 
 		it("defaults unknown model IDs to the OpenAI-compatible format", () => {
diff --git a/src/api/providers/fetchers/__tests__/opencode-go.spec.ts b/src/api/providers/fetchers/__tests__/opencode-go.spec.ts
index 20040ffb22..6d886819b3 100644
--- a/src/api/providers/fetchers/__tests__/opencode-go.spec.ts
+++ b/src/api/providers/fetchers/__tests__/opencode-go.spec.ts
@@ -87,6 +87,38 @@ describe("Opencode Go Fetchers", () => {
 			})
 		})
 
+		it("uses native model info when the endpoint returns only an OpenCode Go model ID", async () => {
+			mockedAxios.get.mockResolvedValue({
+				data: {
+					data: [
+						{ id: "gpt-5.6-luna" },
+						{ id: "hy4-preview" },
+						{ id: "deepseek-v4-flash-vision-exp" },
+						{ id: "muse-spark-1.3-contributor" },
+					],
+				},
+			})
+
+			const models = await getOpencodeGoModels("k")
+
+			expect(models["gpt-5.6-luna"]).toMatchObject({ contextWindow: 1_050_000, maxTokens: 128_000 })
+			expect(models["hy4-preview"]).toMatchObject({ contextWindow: 1_024_000, maxTokens: 64_000 })
+			expect(models["deepseek-v4-flash-vision-exp"]).toMatchObject({
+				contextWindow: 1_000_000,
+				maxTokens: 384_000,
+			})
+			expect(models["muse-spark-1.3-contributor"]).toMatchObject({
+				contextWindow: 1_048_576,
+				maxTokens: 131_072,
+				supportsImages: true,
+				supportsMaxTokens: true,
+				supportsReasoningEffort: ["minimal", "low", "medium", "high", "xhigh"],
+				inputPrice: 0.1,
+				outputPrice: 0.2,
+				cacheReadsPrice: 0.002,
+			})
+		})
+
 		it("returns an empty map on network error", async () => {
 			mockedAxios.get.mockRejectedValue(new Error("network"))
 			expect(await getOpencodeGoModels("k")).toEqual({})
@@ -121,6 +153,59 @@ describe("Opencode Go Fetchers", () => {
 	})
 
 	describe("parseOpencodeGoModel", () => {
+		it("has complete model info for every model returned by OpenCode Go", () => {
+			const modelIds = [
+				"minimax-m3",
+				"minimax-m2.7",
+				"minimax-m2.5",
+				"kimi-k3",
+				"kimi-k2.7-code",
+				"kimi-k2.6",
+				"longcat-2.0",
+				"kimi-k2.5",
+				"glm-5.2",
+				"glm-5.3-flash",
+				"glm-5.3",
+				"glm-5.1",
+				"glm-5",
+				"deepseek-v4-pro",
+				"deepseek-v4-flash",
+				"deepseek-v4-flash-vision-exp",
+				"qwen3.7-max",
+				"qwen3.8-max",
+				"qwen3.8-flash",
+				"qwen3.7-plus",
+				"qwen3.6-plus",
+				"qwen3.5-plus",
+				"mimo-v2-pro",
+				"mimo-v2-omni",
+				"mimo-v2.5-pro",
+				"mimo-v2.5",
+				"hy4-preview",
+				"hy3",
+				"hy3-preview",
+				"gpt-5.6-luna",
+				"grok-4.5",
+				"grok-4.6",
+				"muse-spark-1.3-contributor",
+				"muse-spark-1.2-contributor",
+			]
+			expect(Object.keys(opencodeGoModels).sort()).toEqual([...modelIds].sort())
+
+			for (const modelId of modelIds) {
+				const info = getOpencodeGoModelInfo(modelId)
+				expect(info, modelId).toBeDefined()
+				expect(info?.contextWindow, modelId).toBeGreaterThan(0)
+				expect(info?.maxTokens, modelId).toBeGreaterThan(0)
+				expect(info?.supportsImages, modelId).toEqual(expect.any(Boolean))
+				expect(info?.supportsPromptCache, modelId).toEqual(expect.any(Boolean))
+				expect(info?.inputPrice, modelId).toEqual(expect.any(Number))
+				expect(info?.outputPrice, modelId).toEqual(expect.any(Number))
+				expect(info?.cacheReadsPrice, modelId).toEqual(expect.any(Number))
+				expect(info?.description, modelId).toBeTruthy()
+			}
+		})
+
 		it("merges live endpoint values over the native registry for a curated model", () => {
 			const info = parseOpencodeGoModel({ id: "glm-5.1", context_window: 150000, max_output_tokens: 8000 })
 			// Live values win for volatile fields.
diff --git a/src/api/providers/fetchers/opencode-go.ts b/src/api/providers/fetchers/opencode-go.ts
index d28b8872f0..6b2ad361b2 100644
--- a/src/api/providers/fetchers/opencode-go.ts
+++ b/src/api/providers/fetchers/opencode-go.ts
@@ -41,15 +41,12 @@ const opencodeGoModelsResponseSchema = z.object({
  * slider, and accurate cost reporting.
  *
  * Resolution order for a fully-populated {@link ModelInfo}:
- *   1. Start from the native registry ({@link getOpencodeGoModelInfo}) when the
- *      model ID is curated — this supplies correct context lengths, max tokens,
- *      capability flags, and pricing sourced from vendor specs.
- *   2. Override `contextWindow`, `maxTokens`, and `supportsImages` with values
- *      from the live `/models` payload when present, so the gateway stays the
- *      source of truth for those volatile fields.
- *   3. Fall back to {@link opencodeGoDefaultModelInfo} for any field still
- *      missing on an unknown (non-curated) model, ensuring downstream consumers
- *      always receive a fully-populated object.
+ *   1. Use the native registry ({@link getOpencodeGoModelInfo}) when the model
+ *      is curated, including its capabilities and pricing.
+ *   2. Override static limits and image support with live `/models` values when
+ *      present, keeping the gateway authoritative for volatile fields.
+ *   3. Fall back to {@link opencodeGoDefaultModelInfo} for an unknown model,
+ *      ensuring downstream consumers always receive a fully-populated object.
  *
  * @param model - Validated model entry from the `/models` response.
  * @returns Normalised model metadata suitable for the model picker.
diff --git a/src/api/providers/opencode-go.ts b/src/api/providers/opencode-go.ts
index bbfa63780a..46ea52a2be 100644
--- a/src/api/providers/opencode-go.ts
+++ b/src/api/providers/opencode-go.ts
@@ -690,7 +690,7 @@ export class OpencodeGoHandler extends RouterProvider implements SingleCompletio
 	 * Performs a non-streaming chat completion and returns the full response text.
 	 *
 	 * Anthropic-format models are completed via the `/v1/messages` endpoint;
-	 * Responses-format models (gpt-5.6-luna) via `/v1/responses`; all other
+	 * Responses-format models via `/v1/responses`; all other
 	 * models use the OpenAI-compatible chat completions endpoint.
 	 *
 	 * @param prompt - The user prompt to send as a single user message.