fix(shared): restrict cached Anthropic 1M context to GA 4.6 models only

src/shared/context-limit-resolver.test.ts

@@ -28,7 +28,7 @@ describe("resolveActualContextLimit", () => {
     resetContextLimitEnv()
   })
 
-  it("returns cached limit for Anthropic models when 1M mode is disabled (GA support)", () => {
+  it("returns cached limit for Anthropic 4.6 models when 1M mode is disabled (GA support)", () => {
     // given
     delete process.env[ANTHROPIC_CONTEXT_ENV_KEY]
     delete process.env[VERTEX_CONTEXT_ENV_KEY]
@@ -45,6 +45,23 @@ describe("resolveActualContextLimit", () => {
     expect(actualLimit).toBe(1_000_000)
   })
 
+  it("returns default 200K for older Anthropic models even when cached limit is higher", () => {
+    // given
+    delete process.env[ANTHROPIC_CONTEXT_ENV_KEY]
+    delete process.env[VERTEX_CONTEXT_ENV_KEY]
+    const modelContextLimitsCache = new Map<string, number>()
+    modelContextLimitsCache.set("anthropic/claude-sonnet-4-5", 500_000)
+
+    // when
+    const actualLimit = resolveActualContextLimit("anthropic", "claude-sonnet-4-5", {
+      anthropicContext1MEnabled: false,
+      modelContextLimitsCache,
+    })
+
+    // then
+    expect(actualLimit).toBe(200_000)
+  })
+
   it("returns default 200K for Anthropic models without cached limit and 1M mode disabled", () => {
     // given
     delete process.env[ANTHROPIC_CONTEXT_ENV_KEY]
@@ -92,6 +109,23 @@ describe("resolveActualContextLimit", () => {
     expect(actualLimit).toBe(200000)
   })
 
+  it("supports Anthropic 4.6 dot-version model IDs without explicit 1M mode", () => {
+    // given
+    delete process.env[ANTHROPIC_CONTEXT_ENV_KEY]
+    delete process.env[VERTEX_CONTEXT_ENV_KEY]
+    const modelContextLimitsCache = new Map<string, number>()
+    modelContextLimitsCache.set("anthropic/claude-opus-4.6", 1_000_000)
+
+    // when
+    const actualLimit = resolveActualContextLimit("anthropic", "claude-opus-4.6", {
+      anthropicContext1MEnabled: false,
+      modelContextLimitsCache,
+    })
+
+    // then
+    expect(actualLimit).toBe(1_000_000)
+  })
+
   it("returns null for non-Anthropic providers without a cached limit", () => {
     // given
     delete process.env[ANTHROPIC_CONTEXT_ENV_KEY]

src/shared/context-limit-resolver.ts

@@ -1,6 +1,12 @@
 import process from "node:process"
 
 const DEFAULT_ANTHROPIC_ACTUAL_LIMIT = 200_000
+const ANTHROPIC_NO_HEADER_GA_MODEL_IDS = new Set([
+  "claude-opus-4-6",
+  "claude-opus-4.6",
+  "claude-sonnet-4-6",
+  "claude-sonnet-4.6",
+])
 
 export type ContextLimitModelCacheState = {
   anthropicContext1MEnabled: boolean
@@ -20,6 +26,10 @@ function getAnthropicActualLimit(modelCacheState?: ContextLimitModelCacheState):
     : DEFAULT_ANTHROPIC_ACTUAL_LIMIT
 }
 
+function isAnthropicNoHeaderGaModel(modelID: string): boolean {
+  return ANTHROPIC_NO_HEADER_GA_MODEL_IDS.has(modelID.toLowerCase())
+}
+
 export function resolveActualContextLimit(
   providerID: string,
   modelID: string,
@@ -30,7 +40,7 @@ export function resolveActualContextLimit(
     if (explicit1M === 1_000_000) return explicit1M
 
     const cachedLimit = modelCacheState?.modelContextLimitsCache?.get(`${providerID}/${modelID}`)
-    if (cachedLimit) return cachedLimit
+    if (cachedLimit && isAnthropicNoHeaderGaModel(modelID)) return cachedLimit
 
     return DEFAULT_ANTHROPIC_ACTUAL_LIMIT
   }