Skip to content

A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.

remove old models, add latest ones (#8129)

Shahmir Varqhad85488dc289f
diff --git a/packages/llm-info/README.md b/packages/llm-info/README.md
index a4ee0446e8..8f670d873c 100644
--- a/packages/llm-info/README.md
+++ b/packages/llm-info/README.md
@@ -8,4 +8,8 @@ Models and providers are stored in the `data` directory.
## Adding a New LLM Model
-If you want to add a new LLM model or provider, you can do so by editing the YAML files in the `data` directory (`models.yaml` or `providers.yaml`) and running `pnpm codegen`.
\ No newline at end of file
+If you want to add a new LLM model or provider, you can do so by editing the YAML files in the `data` directory (`models.yml` or `providers.yml`) and running `pnpm codegen`.
+
+> **Note:**
+> To make it easier for users to choose, keep the number of models to a minimum. Focus on including the latest or recommended models from each provider.
+> Review https://openrouter.ai/provider/{provider_name} for the latest models and capabilities.
diff --git a/packages/llm-info/data/models.yml b/packages/llm-info/data/models.yml
index 8fbca7e744..c8d08db268 100644
--- a/packages/llm-info/data/models.yml
+++ b/packages/llm-info/data/models.yml
@@ -1,99 +1,72 @@
-# Information provided by https://hub.continue.dev/
+# Information provided by https://openrouter.ai/
+# Keep latest models at the top, deprecate old models
# Anthropic
-- name: Claude 4.6 Opus
- model: claude-opus-4-6
- description: Latest Opus model with enhanced capabilities and performance improvements
- providers: [anthropic]
- roles: [chat, edit]
- thinking: true
-
-- name: Claude 4.5 Opus
- model: claude-opus-4-5
- description: Modern Opus model with enhanced capabilities and performance improvements
+- name: Claude Sonnet 4.6
+ model: claude-sonnet-4-6
+ description: Most capable Sonnet-class model, with frontier performance across coding, agents, and professional work
providers: [anthropic]
roles: [chat, edit]
thinking: true
-- name: Claude 4.5 Sonnet
- model: claude-sonnet-4-5
- description: Latest Sonnet model with enhanced capabilities and performance improvements
+- name: Claude Opus 4.6
+ model: claude-opus-4-6
+ description: Strongest model for coding and long-running professional tasks
providers: [anthropic]
roles: [chat, edit]
thinking: true
-- name: Claude 4.1 Opus
- model: claude-opus-4-1
- description: Latest Opus model with enhanced capabilities and performance improvements
+- name: Claude Opus 4.5
+ model: claude-opus-4-5
+ description: Frontier reasoning model optimized for complex software engineering
providers: [anthropic]
roles: [chat, edit]
thinking: true
-- name: Claude 4 Opus
- model: claude-opus-4-0
- description: Most capable model with superior performance on complex tasks and reasoning
+- name: Claude Sonnet 4.5
+ model: claude-sonnet-4-5
+ description: Advanced Sonnet model optimized for real-world agents and coding workflows
providers: [anthropic]
roles: [chat, edit]
thinking: true
-- name: Claude 4 Sonnet
- model: claude-sonnet-4-0
- description: Next-generation Sonnet model with advanced reasoning and multimodal capabilities
+- name: Claude Haiku 4.5
+ model: claude-haiku-4-5
+ description: Most efficient model, delivering near-frontier intelligence at a fraction of the cost
providers: [anthropic]
roles: [chat, edit]
thinking: true
-- name: Claude 3.7 Sonnet
- model: claude-3-7-sonnet-latest
- description: Enhanced Sonnet model with improved capabilities
- providers: [anthropic]
- roles: [chat, edit]
- thinking: false
-
-- name: Claude 3.5 Sonnet
- model: claude-3-5-sonnet-latest
- description: Balanced model with strong performance across reasoning, math, and coding
- providers: [anthropic]
- roles: [chat, edit]
- thinking: false
-
-- name: Claude 3.5 Haiku
- model: claude-3-5-haiku-latest
- description: Fast and efficient model with excellent performance for everyday tasks
- providers: [anthropic]
- roles: [chat, edit]
- thinking: false
-
# Google
-- name: Gemini 3.0 Pro
- model: gemini-3-pro-preview
- description: Most capable Gemini model with advanced reasoning and large context support
+- name: Gemini 3.1 Flash Lite Preview
+ model: gemini-3.1-flash-lite-preview
+ description: High-efficiency model optimized for high-volume use cases
providers: [google]
roles: [chat, edit]
- thinking: true
+ thinking: false
-- name: Gemini 2.5 Pro
- model: gemini-2.5-pro
- description: Most capable Gemini model with advanced reasoning and large context support
+- name: Gemini 3.1 Pro Preview
+ model: gemini-3.1-pro-preview
+ description: Frontier reasoning model with enhanced software engineering and agentic reliability
providers: [google]
roles: [chat, edit]
- thinking: false
+ thinking: true
-- name: Gemini 2.5 Pro Experimental
- model: gemini-2.5-pro-exp
- description: Experimental version of Gemini 2.5 Pro with latest improvements
+- name: Gemini 3 Flash Preview
+ model: gemini-3-flash-preview
+ description: High-speed thinking model for agentic workflows, multi-turn chat, and coding assistance
providers: [google]
roles: [chat, edit]
- thinking: false
+ thinking: true
- name: Gemini 2.5 Flash
model: gemini-2.5-flash
- description: Adaptive thinking, cost efficiency
+ description: State-of-the-art workhorse model with built-in thinking capabilities
providers: [google]
roles: [chat, edit]
- thinking: false
+ thinking: true
- name: Gemini 2.5 Flash Lite
model: gemini-2.5-flash-lite
@@ -102,61 +75,33 @@
roles: [chat, edit]
thinking: false
-- name: Gemini 2.0 Flash
- model: gemini-2.0-flash
- description: Next-generation Flash model with improved performance and capabilities
- providers: [google]
- roles: [chat, edit]
- thinking: false
-
-- name: Gemini 2.0 Flash Experimental
- model: gemini-2.0-flash-exp
- description: Experimental version of Gemini 2.0 Flash with cutting-edge features
- providers: [google]
- roles: [chat, edit]
- thinking: false
+# OpenAI
-- name: Gemini 2.0 Flash Lite
- model: gemini-2.0-flash-lite
- description: Lightweight version of Gemini 2.0 Flash optimized for efficiency
- providers: [google]
+- name: GPT-5.3 Chat
+ model: gpt-5.3-chat
+ description: Most-used model that makes everyday conversations smoother and more useful
+ providers: [openai, azure]
roles: [chat, edit]
thinking: false
-- name: Gemini 2.0 Flash Thinking
- model: gemini-2.0-flash-thinking-exp-01-21
- description: Specialized reasoning model with enhanced thinking capabilities
- providers: [google]
- roles: [chat]
- thinking: true
-
-- name: Gemini 1.5 Flash
- model: gemini-1.5-flash
- description: Fast and efficient model optimized for speed and everyday tasks
- providers: [google]
+- name: GPT-5.3-Codex
+ model: gpt-5.3-codex
+ description: Most advanced agentic coding model combining frontier software engineering with broad reasoning
+ providers: [openai]
roles: [chat, edit]
thinking: false
-- name: Gemini 1.5 Flash 8B
- model: gemini-1.5-flash-8b
- description: Lightweight version of Gemini 1.5 Flash with 8 billion parameters
- providers: [google]
- roles: [chat, edit]
- thinking: false
-
-- name: Gemini 1.5 Pro
- model: gemini-1.5-pro
- description: Advanced model with superior reasoning capabilities and large context window
- providers: [google]
+- name: GPT-5.2
+ model: gpt-5.2
+ description: Latest frontier-grade model with strong agentic and long context performance
+ providers: [openai, azure]
roles: [chat, edit]
thinking: false
-# OpenAI
-
-- name: GPT-5
- model: gpt-5
- description: The best model for coding and agentic tasks across domains
- providers: [openai, azure, github]
+- name: GPT-5.2 Pro
+ model: gpt-5.2-pro
+ description: Powerful model for complex coding tasks, long-form writing, and advanced reasoning
+ providers: [openai]
roles: [chat, edit]
thinking: false
@@ -174,51 +119,9 @@
roles: [chat, edit]
thinking: false
-- name: GPT-4.1
- model: gpt-4.1
- description: Fast, highly intelligent model with largest context window (1 million tokens)
- providers: [openai, azure]
- roles: [chat, edit]
- thinking: false
-
-- name: GPT-4.1 Mini
- model: gpt-4.1-mini
- description: Balanced for intelligence, speed, and cost
- providers: [openai, azure]
- roles: [chat, edit]
- thinking: false
-
-- name: GPT-4o
- model: gpt-4o
- description: Fast, intelligent, flexible GPT model with multimodal capabilities
- providers: [openai, azure]
- roles: [chat, edit]
- thinking: false
-
-- name: GPT-4o Mini
- model: gpt-4o-mini
- description: Fast, affordable small model for focused tasks
- providers: [openai, azure, github]
- roles: [chat, edit]
- thinking: false
-
- name: o3
model: o3
- description: Our most powerful reasoning model with chain-of-thought capabilities
- providers: [openai, azure]
- roles: [chat, edit]
- thinking: true
-
-- name: o3 Mini
- model: o3-mini
- description: A small model alternative to o3 for reasoning tasks
- providers: [openai, azure]
- roles: [chat, edit]
- thinking: true
-
-- name: o3 Pro
- model: o3-pro
- description: Version of o3 designed to think longer and provide most reliable responses
+ description: Most powerful reasoning model with chain-of-thought capabilities
providers: [openai, azure]
roles: [chat, edit]
thinking: true
@@ -241,40 +144,33 @@
# Bedrock
-- name: Claude 4.5 Sonnet (Bedrock)
- model: global.anthropic.claude-sonnet-4-5-20250929-v1:0
- description: Enhanced Sonnet model with improved capabilities
- providers: [bedrock]
- roles: [chat, edit]
- thinking: false
-
-- name: Claude 4.1 Opus (Bedrock)
- model: us.anthropic.claude-opus-4-1-20250805-v1:0
- description: Opus model for deep thinking
+- name: Claude 4.6 Opus (Bedrock)
+ model: global.anthropic.claude-opus-4-6-20260206-v1:0
+ description: Strongest model for coding and long-running professional tasks
providers: [bedrock]
roles: [chat, edit]
- thinking: false
+ thinking: true
-- name: Claude 3.7 Sonnet (Bedrock)
- model: us.anthropic.claude-3-7-sonnet-20250219-v1:0
- description: Enhanced Sonnet model with improved capabilities
+- name: Claude 4.5 Opus (Bedrock)
+ model: global.anthropic.claude-opus-4-5-20251120-v1:0
+ description: Frontier reasoning model optimized for complex software engineering
providers: [bedrock]
roles: [chat, edit]
- thinking: false
+ thinking: true
-- name: Claude 3.5 Sonnet v2 (Bedrock)
- model: us.anthropic.claude-3-5-sonnet-20241022-v2:0
- description: Balanced model with strong performance across reasoning, math, and coding
+- name: Claude 4.5 Sonnet (Bedrock)
+ model: global.anthropic.claude-sonnet-4-5-20250929-v1:0
+ description: Most advanced Sonnet model to date, optimized for real-world agents
providers: [bedrock]
roles: [chat, edit]
- thinking: false
+ thinking: true
-- name: Claude 3.5 Haiku (Bedrock)
- model: us.anthropic.claude-3-5-haiku-20241022-v1:0
- description: Fast and efficient model with excellent performance for everyday tasks
+- name: Claude 4.5 Haiku (Bedrock)
+ model: global.anthropic.claude-haiku-4-5-latest-v1:0
+ description: Most efficient model, delivering near-frontier intelligence at a fraction of the cost
providers: [bedrock]
roles: [chat, edit]
- thinking: false
+ thinking: true
- name: Llama 3.3 70B Instruct
model: us.meta.llama3-3-70b-instruct-v1:0
@@ -301,7 +197,7 @@
- name: gpt-oss 20b
model: gpt-oss:20b
- description: OpenAI’s open-weight models designed for powerful reasoning, agentic tasks, and versatile developer use cases.
+ description: OpenAI's open-weight models designed for powerful reasoning, agentic tasks, and versatile developer use cases.
providers: [ollama]
roles: [chat, edit]
thinking: true
diff --git a/packages/llm-info/src/generate.ts b/packages/llm-info/src/generate.ts
index af4b8bc7f3..bc8ab7dcc3 100644
--- a/packages/llm-info/src/generate.ts
+++ b/packages/llm-info/src/generate.ts
@@ -7,7 +7,7 @@ import { parse } from "yaml";
import { z } from "zod";
import { Logger } from "./simple_logger.ts";
-const ROLES = ["chat", "edit", "rerank", "embed"] as const;
+const ROLES = ["chat", "edit", "rerank", "embed", "autocomplete"] as const;
export const LLMInfoSchema = z.object({
name: z.string(),