diff --git a/CHANGELOG.md b/CHANGELOG.md
index e19554671..304505ec0 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,7 @@
 - Onboarding/Models: add first-class Z.AI (GLM) auth choice (`zai-api-key`) + `--zai-api-key` flag.
 - Agents: add human-delay pacing between block replies (modes: off/natural/custom, per-agent configurable). (#446) — thanks @tony-freedomology.
 - Onboarding/Models: add catalog-backed default model picker to onboarding + configure. (#611) — thanks @jonasjancarik.
+- Agents/OpenCode Zen: update fallback models + defaults, keep legacy alias mappings. (#669) — thanks @magimetal.
 
 ### Fixes
 - Auto-reply: prefer `RawBody` for command/directive parsing (WhatsApp + Discord) and prevent fallback runs from clobbering concurrent session updates. (#643) — thanks @mcinteerj.
diff --git a/src/agents/opencode-zen-models.test.ts b/src/agents/opencode-zen-models.test.ts
index 208ac18bb..4fcbe3395 100644
--- a/src/agents/opencode-zen-models.test.ts
+++ b/src/agents/opencode-zen-models.test.ts
@@ -12,6 +12,14 @@ describe("resolveOpencodeZenAlias", () => {
     expect(resolveOpencodeZenAlias("opus")).toBe("claude-opus-4-5");
   });
 
+  it("keeps legacy aliases working", () => {
+    expect(resolveOpencodeZenAlias("sonnet")).toBe("claude-opus-4-5");
+    expect(resolveOpencodeZenAlias("haiku")).toBe("claude-opus-4-5");
+    expect(resolveOpencodeZenAlias("gpt4")).toBe("gpt-5.1");
+    expect(resolveOpencodeZenAlias("o1")).toBe("gpt-5.2");
+    expect(resolveOpencodeZenAlias("gemini-2.5")).toBe("gemini-3-pro");
+  });
+
   it("resolves gpt5 alias", () => {
     expect(resolveOpencodeZenAlias("gpt5")).toBe("gpt-5.2");
   });
@@ -88,5 +96,12 @@ describe("OPENCODE_ZEN_MODEL_ALIASES", () => {
     expect(OPENCODE_ZEN_MODEL_ALIASES.gemini).toBe("gemini-3-pro");
     expect(OPENCODE_ZEN_MODEL_ALIASES.glm).toBe("glm-4.7-free");
     expect(OPENCODE_ZEN_MODEL_ALIASES.minimax).toBe("minimax-m2.1-free");
+
+    // Legacy aliases (kept for backward compatibility).
+    expect(OPENCODE_ZEN_MODEL_ALIASES.sonnet).toBe("claude-opus-4-5");
+    expect(OPENCODE_ZEN_MODEL_ALIASES.haiku).toBe("claude-opus-4-5");
+    expect(OPENCODE_ZEN_MODEL_ALIASES.gpt4).toBe("gpt-5.1");
+    expect(OPENCODE_ZEN_MODEL_ALIASES.o1).toBe("gpt-5.2");
+    expect(OPENCODE_ZEN_MODEL_ALIASES["gemini-2.5"]).toBe("gemini-3-pro");
   });
 });
diff --git a/src/agents/opencode-zen-models.ts b/src/agents/opencode-zen-models.ts
index 4df00c4d9..3cdcd0753 100644
--- a/src/agents/opencode-zen-models.ts
+++ b/src/agents/opencode-zen-models.ts
@@ -27,12 +27,29 @@ export const OPENCODE_ZEN_MODEL_ALIASES: Record<string, string> = {
   // Claude
   opus: "claude-opus-4-5",
   "opus-4.5": "claude-opus-4-5",
+  "opus-4": "claude-opus-4-5",
+
+  // Legacy Claude aliases (OpenCode Zen rotates model catalogs; keep old keys working).
+  sonnet: "claude-opus-4-5",
+  "sonnet-4": "claude-opus-4-5",
+  haiku: "claude-opus-4-5",
+  "haiku-3.5": "claude-opus-4-5",
 
   // GPT-5.x family
   gpt5: "gpt-5.2",
   "gpt-5": "gpt-5.2",
   "gpt-5.1": "gpt-5.1",
 
+  // Legacy GPT aliases (keep old config/docs stable; map to closest current equivalents).
+  gpt4: "gpt-5.1",
+  "gpt-4": "gpt-5.1",
+  "gpt-mini": "gpt-5.1-codex-mini",
+
+  // Legacy O-series aliases (no longer in the Zen catalog; map to a strong default).
+  o1: "gpt-5.2",
+  o3: "gpt-5.2",
+  "o3-mini": "gpt-5.1-codex-mini",
+
   // Codex family
   codex: "gpt-5.1-codex",
   "codex-mini": "gpt-5.1-codex-mini",
@@ -45,6 +62,11 @@ export const OPENCODE_ZEN_MODEL_ALIASES: Record<string, string> = {
   flash: "gemini-3-flash",
   "gemini-flash": "gemini-3-flash",
 
+  // Legacy Gemini 2.5 aliases (map to the nearest current Gemini tier).
+  "gemini-2.5": "gemini-3-pro",
+  "gemini-2.5-pro": "gemini-3-pro",
+  "gemini-2.5-flash": "gemini-3-flash",
+
   // GLM (free + alpha)
   glm: "glm-4.7-free",
   "glm-free": "glm-4.7-free",
@@ -161,6 +183,7 @@ function buildModelDefinition(modelId: string): ModelDefinitionConfig {
     id: modelId,
     name: formatModelName(modelId),
     api: resolveOpencodeZenModelApi(modelId),
+    // Treat Zen models as reasoning-capable so defaults pick thinkLevel="low" unless users opt out.
     reasoning: true,
     input: supportsImageInput(modelId) ? ["text", "image"] : ["text"],
     cost: MODEL_COSTS[modelId] ?? DEFAULT_COST,