From 53f6168954c00f3231339931d5316727cfba4a85 Mon Sep 17 00:00:00 2001
From: Ankur Goyal <ankrgyl@gmail.com>
Date: Thu, 1 Oct 2026 21:11:57 +0000
Subject: [PATCH] Use GPT-6 Luna as default model

---
 js/llm.test.ts           |  5 +++++
 js/llm.ts                |  2 +-
 py/autoevals/llm.py      |  2 +-
 py/autoevals/test_llm.py | 13 ++++++++++++-
 4 files changed, 19 insertions(+), 3 deletions(-)

diff --git a/js/llm.test.ts b/js/llm.test.ts
index 5863eecc..d5d4fc0d 100644
--- a/js/llm.test.ts
+++ b/js/llm.test.ts
@@ -6,6 +6,7 @@ import { afterAll, afterEach, beforeAll, describe, expect, test } from "vitest";
 import {
   Battle,
   buildClassificationTools,
+  DEFAULT_MODEL,
   LLMClassifierFromTemplate,
   OpenAIClassifier,
   templateUsesThreadVariables,
@@ -65,6 +66,10 @@ afterAll(() => {
 });
 
 describe("LLM Tests", () => {
+  test("exports the default scorer model", () => {
+    expect(DEFAULT_MODEL).toBe("gpt-6-luna");
+  });
+
   test("templateUsesThreadVariables recognizes thread_with_system", () => {
     expect(templateUsesThreadVariables("{{thread_with_system}}")).toBe(true);
     expect(
diff --git a/js/llm.ts b/js/llm.ts
index 1b7f318e..1e572324 100644
--- a/js/llm.ts
+++ b/js/llm.ts
@@ -86,7 +86,7 @@ export type LLMArgs = {
  * The default model to use for LLM-based evaluations.
  * @deprecated Use `init({ defaultModel: "..." })` to configure the default model instead.
  */
-export const DEFAULT_MODEL = "gpt-5-mini";
+export const DEFAULT_MODEL = "gpt-6-luna";
 
 const PLAIN_RESPONSE_SCHEMA = {
   properties: {
diff --git a/py/autoevals/llm.py b/py/autoevals/llm.py
index b0093cb4..1a38712a 100644
--- a/py/autoevals/llm.py
+++ b/py/autoevals/llm.py
@@ -88,7 +88,7 @@
 )
 
 # Deprecated: Use init(default_model="...") to configure the default model instead.
-DEFAULT_MODEL = "gpt-5-mini"
+DEFAULT_MODEL = "gpt-6-luna"
 
 PLAIN_RESPONSE_SCHEMA = {
     "properties": {"choice": {"description": "The choice", "title": "Choice", "type": "string"}},
diff --git a/py/autoevals/test_llm.py b/py/autoevals/test_llm.py
index 8e987efa..2bcecefc 100644
--- a/py/autoevals/test_llm.py
+++ b/py/autoevals/test_llm.py
@@ -9,7 +9,14 @@
 from pydantic import BaseModel
 
 from autoevals import init
-from autoevals.llm import Battle, Factuality, LLMClassifier, OpenAILLMClassifier, build_classification_tools
+from autoevals.llm import (
+    DEFAULT_MODEL,
+    Battle,
+    Factuality,
+    LLMClassifier,
+    OpenAILLMClassifier,
+    build_classification_tools,
+)
 from autoevals.oai import OpenAIV1Module, get_default_model
 from autoevals.thread_utils import compute_thread_template_vars, template_uses_thread_variables
 
@@ -19,6 +26,10 @@ class TestModel(BaseModel):
     num: int
 
 
+def test_default_scorer_model():
+    assert DEFAULT_MODEL == "gpt-6-luna"
+
+
 def test_render_messages():
     classifier = OpenAILLMClassifier(
         "test",