---
id: post/thinkaboutit-be/8185
type: post
title: Choosing a Language Model for AL Development in VS Code
summary: When using Copilot Chat in VS Code for AL development, you can select different language models based on task complexity. Fast models suit simple tasks while reasoning models handle design decisions, but AL developers must validate model suggestions against actual BC symbols since models may confidently produce outdated code based on old training data.
tier: community
language: en
tags:
  - language models
  - copilot chat
  - al development
  - vs code
  - ai assistance
  - code generation
system: development
review:
  state: reviewed
  by: opus
  at: "2026-10-07T23:46:27.148Z"
  flags: []
generated:
  at: "2026-10-07T22:42:12.892Z"
  pipeline: 0.2.0
  prompts:
    extract-post: 1
  input_hash: 4a42de38992df9b85085a6534d7d43f420db6759914d6be814a4230d777bfe5e
evidence:
  - kind: blog
    url: https://thinkaboutit.be/2026/09/choosing-a-language-model-for-al-development-in-vs-code/
    title: Choosing a Language Model for AL Development in VS Code
    date: "2026-09-23"
    commit: null
    t: null
    quote: null
  - kind: blog
    url: https://thinkaboutit.be/2026/09/choosing-a-language-model-for-al-development-in-vs-code/
    title: Choosing a Language Model for AL Development in VS Code
    date: "2026-09-23"
    commit: null
    t: null
    quote: A model trained before a given BC version shipped will write code with full confidence, using an older version of the standard application.
  - kind: blog
    url: https://thinkaboutit.be/2026/09/choosing-a-language-model-for-al-development-in-vs-code/
    title: Choosing a Language Model for AL Development in VS Code
    date: "2026-09-23"
    commit: null
    t: null
    quote: Whenever you write AL code against a standard Business Central object, treat the model's first answer as a guess, not a fact.
links:
  learn: []
  objects: []
  features: []
  topics: []
  localizations: []
  videos: []
  posts: []
  guidelines: []
post_id: "8185"
source_id: thinkaboutit-be
source_name: think about IT - Steven Renders
url: https://thinkaboutit.be/2026/09/choosing-a-language-model-for-al-development-in-vs-code/
published_at: "2026-09-23T12:06:18.000Z"
author: Steven Renders
full_text: false
words: 950
quotes:
  - text: A model trained before a given BC version shipped will write code with full confidence, using an older version of the standard application.
    why_it_matters: "Highlights the core AL-specific problem: models confidently generate outdated code without warning, which is more dangerous than general coding mistakes"
  - text: Whenever you write AL code against a standard Business Central object, treat the model's first answer as a guess, not a fact.
    why_it_matters: Provides essential guidance for AL developers to validate model output against real symbols to catch knowledge gaps
code_objects_mentioned: []
systems:
  - development
  - copilot
  - platform
versions_mentioned: []
preview:
  embeddable: true
  frame_url: https://thinkaboutit.be/2026/09/choosing-a-language-model-for-al-development-in-vs-code/embed/
  image: https://i0.wp.com/thinkaboutit.be/wp-content/uploads/2026/09/2026-09-23_13-53-10.png?fit=301%2C207&ssl=1
  image_alt: null
  image_w: 301
  image_h: 207
  site_name: think about IT
  favicon: https://i0.wp.com/thinkaboutit.be/wp-content/uploads/2026/02/cropped-SiteIcon2-1.png?fit=32%2C32&ssl=1
  probed_at: "2026-10-07T11:50:06.396Z"
---

# Choosing a Language Model for AL Development in VS Code

[Read the post](https://thinkaboutit.be/2026/09/choosing-a-language-model-for-al-development-in-vs-code/) · think about IT - Steven Renders (Steven Renders) · 2026-09-23 · 950 words · tier community · reviewed (checked by Opus)

> When using Copilot Chat in VS Code for AL development, you can select different language models based on task complexity. Fast models suit simple tasks while reasoning models handle design decisions, but AL developers must validate model suggestions against actual BC symbols since models may confidently produce outdated code based on old training data.

## Key points

- Language models have training cutoffs and lack knowledge of BC versions released after their training, causing them to suggest outdated code with false confidence
- Fast models work best for simple repetitive AL tasks like standard validations, while reasoning models should be used for design decisions and complex problems
- Model choice applies uniformly across Ask, Plan, and Agent roles in VS Code, though custom agents can pin specific models in their frontmatter
- Always verify model suggestions against your actual BC symbols and objects rather than trusting model output alone
- Context window size and cost vary by model; check the Manage models view for current options and capabilities

## Quotes

- "A model trained before a given BC version shipped will write code with full confidence, using an older version of the standard application." (Highlights the core AL-specific problem: models confidently generate outdated code without warning, which is more dangerous than general coding mistakes)
- "Whenever you write AL code against a standard Business Central object, treat the model's first answer as a guess, not a fact." (Provides essential guidance for AL developers to validate model output against real symbols to catch knowledge gaps)

## Context

- Features: Copilot Chat, model selection, reasoning models, fast models, custom agents, Session Target

Source: think about IT - Steven Renders, community blog. Summary, key points and quotes are derived (CONTENT-NOTICE.md); read the original for the full text.