Purpose guide · Coding

Which AI models belong on a coding shortlist?

Catalog current through July 14, 2026

5 models

Five source-backed models whose Epoch records explicitly include code generation. This is a shortlist, not a performance ranking.

Selection rule

Included when the Epoch record explicitly lists “Code generation.” Release date, lab, access, and weight availability do not improve a model’s position because this page has no positions.

Task-fit shortlist

Evidence first. Your evaluation next.

Ordered by release date for readability, newest first. The order is not a score or recommendation.

  1. 01
    xAI

    Grok 4.1

    Released November 17, 2025

    Matching evidence
    Code generation
    Domain
    Language
    Access
    API access; closed weights
    Model source
  2. 02
    Anthropic

    Claude Haiku 4.5

    Released October 15, 2025

    Matching evidence
    Code generation
    Domain
    Language
    Access
    API access; closed weights
    Model source
  3. 03
    DeepSeek

    DeepSeek-R1 (May 2025)

    Released May 28, 2025

    Matching evidence
    Code generation
    Domain
    Language
    Access
    Open weights (unrestricted); open weights
    Model source
  4. 04
    Meta

    Llama 4 Scout

    Released April 5, 2025

    Matching evidence
    Code generation
    Domain
    Multimodal · Language · Vision
    Access
    Open weights (restricted use); open weights
    Model source
  5. 05
    Meta

    Llama 4 Maverick

    Released April 5, 2025

    Matching evidence
    Code generation
    Domain
    Multimodal · Language · Vision
    Access
    Open weights (restricted use); open weights
    Model source

Decision criteria

Questions the source can help you ask.

These checks narrow an evaluation plan. They do not replace hands-on testing.

01

Do you need downloadable weights?

Use the weight field to separate API-only models from downloadable options. Read the stated use restriction before deploying an open-weight model.

02

Will the work include images?

Domain labels can identify models with vision or multimodal coverage. They do not prove performance on screenshots, diagrams, or interface work.

03

What must you test yourself?

Run representative repository tasks for correctness, tool use, latency, context handling, and cost. Epoch's catalog does not report those fields.

Evidence boundary

What this guide does not know.

No task-level quality, price, context window, latency, or reliability measurement is present in the Epoch catalog. We therefore publish no ranking or “best” label.

Catalog source: Epoch AI Data on AI Models, retrieved 2026-07-14, CC-BY. Artifact sha256:3c663b17ec989f00ef5f994214b878a914118e2d1fb6077e23329a318209eb86.

Each listed model links to its model-level source above. See the publication source ledger and methodology for the wider evidence policy.