Skip to content

An AI-powered search engine with a generative UI

feat(models): upgrade cloud deployment to GPT-6 Luna (#1040)

Yoshiki Miurab58ef319489a
diff --git a/config/models/cloud.json b/config/models/cloud.json
index 60d41326..c446d103 100644
--- a/config/models/cloud.json
+++ b/config/models/cloud.json
@@ -2,8 +2,8 @@
"version": 2,
"models": {
"quick": {
- "id": "gpt-5.6-luna",
- "name": "GPT-5.6 Luna",
+ "id": "gpt-6-luna",
+ "name": "GPT-6 Luna",
"provider": "OpenAI",
"providerId": "openai",
"providerOptions": {
@@ -13,8 +13,8 @@
}
},
"adaptive": {
- "id": "gpt-5.6-luna",
- "name": "GPT-5.6 Luna",
+ "id": "gpt-6-luna",
+ "name": "GPT-6 Luna",
"provider": "OpenAI",
"providerId": "openai",
"providerOptions": {
diff --git a/lib/agents/__tests__/title-generator.test.ts b/lib/agents/__tests__/title-generator.test.ts
index e191d0c0..65580211 100644
--- a/lib/agents/__tests__/title-generator.test.ts
+++ b/lib/agents/__tests__/title-generator.test.ts
@@ -23,11 +23,11 @@ describe('generateChatTitle', () => {
} as never)
})
- it('disables reasoning without limiting output for GPT-5.6 Luna', async () => {
+ it('disables reasoning without limiting output for GPT-6 Luna', async () => {
await expect(
generateChatTitle({
userMessageContent: 'Explain the latest model changes',
- modelId: 'openai:gpt-5.6-luna'
+ modelId: 'openai:gpt-6-luna'
})
).resolves.toBe('Concise title')
diff --git a/lib/agents/title-generator.ts b/lib/agents/title-generator.ts
index ef6c123c..a3847c7d 100644
--- a/lib/agents/title-generator.ts
+++ b/lib/agents/title-generator.ts
@@ -10,7 +10,7 @@ interface GenerateChatTitleParams {
}
function getTitleProviderOptions(modelId: string) {
- if (modelId === 'openai:gpt-5.6-luna') {
+ if (modelId === 'openai:gpt-5.6-luna' || modelId === 'openai:gpt-6-luna') {
return {
openai: {
reasoningEffort: 'none' as const
diff --git a/lib/config/model-metadata.json b/lib/config/model-metadata.json
index b188f4be..165535e2 100644
--- a/lib/config/model-metadata.json
+++ b/lib/config/model-metadata.json
@@ -373,6 +373,11 @@
"input": 922000,
"output": 128000
},
+ "gpt-6-luna": {
+ "context": 1050000,
+ "input": 922000,
+ "output": 128000
+ },
"gpt-image-1": {
"context": 0,
"input": 0,
diff --git a/lib/streaming/helpers/__tests__/trim-cold-start-history.test.ts b/lib/streaming/helpers/__tests__/trim-cold-start-history.test.ts
index 657e1ffb..077844b3 100644
--- a/lib/streaming/helpers/__tests__/trim-cold-start-history.test.ts
+++ b/lib/streaming/helpers/__tests__/trim-cold-start-history.test.ts
@@ -265,7 +265,7 @@ describe('trimColdStartHistory', () => {
).toBe(true)
}, 10_000)
- it('counts gpt-5.6-luna history with its own tokenizer', () => {
+ it('counts gpt-6-luna history with its own tokenizer', () => {
const arabicText = 'مرحبا بك في هذا البحث'.repeat(20)
const coldAt = timestamp(3 * WARM_GAP_MS + COLD_GAP_MS)
const messages = Array.from({ length: 5 }, (_, i) => {
@@ -284,7 +284,7 @@ describe('trimColdStartHistory', () => {
.trimmedAtCurrentTurn
).toBe(true)
expect(
- trimColdStartHistory(messages, { ...options, modelId: 'gpt-5.6-luna' })
+ trimColdStartHistory(messages, { ...options, modelId: 'gpt-6-luna' })
.messages
).toBe(messages)
})
diff --git a/lib/utils/__tests__/context-window.test.ts b/lib/utils/__tests__/context-window.test.ts
index 1e12dd9e..1a03148a 100644
--- a/lib/utils/__tests__/context-window.test.ts
+++ b/lib/utils/__tests__/context-window.test.ts
@@ -13,9 +13,9 @@ import {
describe('context-window', () => {
describe('countTextTokens', () => {
- test('counts gpt-5.6-luna text with the o200k tokenizer', () => {
+ test('counts gpt-6-luna text with the o200k tokenizer', () => {
const text = 'مرحبا بك في هذا البحث عن الطاقة المتجددة. '.repeat(50)
- const tokens = countTextTokens(text, 'gpt-5.6-luna')
+ const tokens = countTextTokens(text, 'gpt-6-luna')
expect(tokens).toBe(getEncoding('o200k_base').encode(text).length)
expect(tokens).toBeLessThan(
@@ -80,11 +80,11 @@ describe('context-window', () => {
}
})
- test('uses the real 1.05M window for GPT-5.6 Luna', () => {
+ test('uses the real 1.05M window for GPT-6 Luna', () => {
// (1050000 - 128000) - floor(1050000 * 0.1) = 817000
const maxTokens = getMaxAllowedTokens({
...mockModel,
- id: 'gpt-5.6-luna'
+ id: 'gpt-6-luna'
})
expect(maxTokens).toBe(817000)
})
diff --git a/lib/utils/context-window.ts b/lib/utils/context-window.ts
index 3efa4b83..08e60745 100644
--- a/lib/utils/context-window.ts
+++ b/lib/utils/context-window.ts
@@ -44,6 +44,7 @@ const MODEL_TO_ENCODING: Record<string, TiktokenEncoding> = {
'gpt-4.1-nano': 'cl100k_base',
'gpt-4o-mini': 'cl100k_base',
'gpt-5.6-luna': 'o200k_base',
+ 'gpt-6-luna': 'o200k_base',
'claude-opus-4': 'cl100k_base', // Use GPT-4 tokenizer as approximation for Claude
'claude-sonnet-4': 'cl100k_base',
'claude-3-7-sonnet': 'cl100k_base',