InCoder
Open-source code infilling model — generates code in the middle of existing functions, not just at the end.
Overview
InCoder is a generative model for code, developed by Facebook AI Research (FAIR) and released in 2022, that was specifically designed for code infilling — generating code that fills a gap in the middle of existing code, given both what comes before and after the blank. This 'fill in the middle' capability was novel at the time and addressed a major limitation of left-to-right code generation models.
The model was trained on a mixture of code from GitHub and GitLab in 28 programming languages, and supports generation, editing, and documentation generation in addition to infilling. The architecture uses a causal masking approach that trains the model to understand code in bidirectional context, enabling coherent gap completion.
InCoder influenced the design of subsequent code models, and the infill training methodology was adopted by Salesforce's CodeGen, Phi, and others. For IDE integration tasks where the cursor is positioned in the middle of a function rather than at the end, InCoder's approach is more suitable than pure left-to-right models.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 8 May 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
Use cases
ASEAN Perspective
InCoder in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
InCoder is a Meta AI research model for code generation and infilling, released as open weights on Hugging Face, notable historically for bidirectional infilling rather than left-to-right completion. It suits researchers and ML engineers studying code models or needing a self-hostable baseline, not developers looking for a day-to-day coding assistant.
The honest caveat is that InCoder is an older, relatively small research artifact long surpassed by Code Llama, StarCoder, DeepSeek-Coder, and current commercial assistants on both quality and tooling. There is no hosted product, no polished UX, and no support; you bring your own inference. Useful for research and reproducibility, not for production coding.
Notable facts
- InCoder was the first large code model to demonstrate that left-to-right generation was not the optimal paradigm for real-world code editing tasks.
- The model achieved state-of-the-art fill-in-the-middle scores on the HumanEval infilling benchmark at its release in 2022.
- Facebook AI Research released InCoder months before GitHub Copilot's general availability, when the concept of AI code completion was still novel to most developers.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including InCoder's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with InCoder unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to InCoder directly →
Spotted something out of date? Suggest an update →
Alternatives to InCoder
More in Code & Dev Tools
