Biome v2 Ships Type-Aware JavaScript Linting Without the TypeScript Compiler

Overview

Biome, the Rust-based web toolchain that has positioned itself as a unified replacement for ESLint and Prettier, shipped its v2.0.0 release on June 17, 2025, according to the project’s GitHub releases page. The major release introduced what the project describes as the first type-aware linting for JavaScript and TypeScript that does not rely on the TypeScript compiler, alongside a new plugin system built on the GritQL query language. By February 2026, the project had iterated to v2.4.0, with v2.4.15 shipping on May 9, 2026, per the GitHub releases page.

What We Know

Type-Aware Linting Without tsc

The central technical claim of Biome v2 is that it can apply type-aware lint rules — the kind that require understanding how values flow between files — without installing the TypeScript package or invoking the TypeScript compiler. InfoQ reported on the beta announcement in June 2025, noting that the first type-aware rule shipped in v2 is noFloatingPromises, which catches async calls whose returned promises are never awaited or handled.

The approach uses Biome’s own type inference engine rather than delegating to tsc, keeping the linter fast. The project seeded the v2 beta as early as March 24, 2025, per the GitHub releases page, before the stable release arrived roughly three months later.

Plugin System via GritQL

Alongside type inference, Biome v2 introduced a plugin system that lets teams write custom lint rules using GritQL, a code query language. Plugins allow users to match code patterns in Biome’s concrete syntax tree and attach custom diagnostic messages. InfoQ noted that the Biome team described the initial plugin scope as limited but said they had “plenty of ideas for making them more powerful.”

Domains: Framework-Aware Rule Organization

Biome v2 also introduced a concept called domains, which groups lint rules by technology or framework, as reported by InfoQ. The initial domains — next, react, solid, and test — allow Biome to auto-detect from a project’s package.json which rule sets are relevant and apply only those, reducing noise in projects that use a single framework.

Rule Coverage and Performance

The Biome repository on GitHub lists more than 450 lint rules, drawn from ESLint, typescript-eslint, and other sources. The project claims 97% compatibility with Prettier for formatting. InfoQ cited a performance figure of 35x faster than Prettier. The GitHub repository shows 24.7k stars and reports the project is licensed under MIT or Apache 2.0.

The Competitive Landscape

The JavaScript linting space has become crowded with Rust-based contenders. ESLint, the dominant incumbent with more than 120 million weekly npm downloads, shipped its own major release — ESLint v10 — in 2026, completing the transition to its flat config system and dropping support for legacy .eslintrc files, according to InfoQ. That same InfoQ article characterized Biome as offering “~95% rule coverage” and noted a second Rust-based challenger, Oxlint, which claims 50 to 100 times faster performance depending on CPU cores but has not yet matched full ESLint rule coverage.

The pressure on ESLint from both Biome and Oxlint was visible in community discussions around ESLint v10. InfoQ reported that migration friction from the legacy config system drew community criticism, with Reddit commenters quoted in the article describing the migration path as poor despite the flat config concept being sound in principle.

v2.4 and Ongoing Development

Biome v2.4.0 arrived on February 15, 2026, per the GitHub releases page. Patch releases continued through v2.4.15 on May 9, 2026, indicating an active maintenance cadence. The project’s Rust-first implementation — GitHub reports 86.3% of the codebase written in Rust — underpins the performance claims across formatting, linting, and the type inference engine.

Biome traces its lineage to Rome, a JavaScript toolchain project that was forked and subsequently renamed to Biome, as noted by InfoQ.

What We Don’t Know

Biome v2’s type inference covers a subset of what the TypeScript compiler handles. The project has not claimed full parity with typescript-eslint’s type-checked rules, and the breadth of edge cases caught by its inference engine compared to tsc-backed alternatives remains a subject of developer evaluation rather than settled benchmarking.

Adoption figures are not independently verifiable through allowlisted sources, making it difficult to assess how widely v2’s type-aware rules are being used in production relative to the larger ESLint ecosystem’s 120 million weekly downloads.

The plugin system, described at launch as a first iteration with limited scope, has not yet been characterized by any independent assessment of how well GritQL covers real-world rule authoring needs compared to ESLint’s JavaScript-native plugin model.

Analysis

Biome v2’s most consequential claim is architectural: type-aware linting has historically required tight coupling to the TypeScript compiler, making it slow and dependent on a working tsc installation. By building its own inference engine, Biome is attempting to break that dependency and offer a path where a single Rust binary handles formatting, linting, and type-aware analysis with no Node.js, TypeScript, or plugin ecosystem required beyond the GritQL layer.

Whether that narrower type model is sufficient for most teams is the practical question the community is now answering. The v2.4 release cycle’s rapid patch cadence and the active GitHub issue tracker suggest the project is iterating quickly on the gaps between its coverage and the full TypeScript type system. Meanwhile, ESLint v10’s flat config transition — criticized for its migration friction — may accelerate evaluation of alternatives in teams already facing configuration churn.