ReactBench v1 (ReactBench)
A coding-agent benchmark for realistic React work, with rubrics that check production concerns such as performance, accessibility, correctness, and code quality.
How we show ReactBench
We mirror the official ReactBench table from Million: 54 effort variants across 39 React tasks. The source score is a weighted rubric aggregate, and solutions that miss a blocking criterion receive zero.
ReactBench tests production React work that ordinary behavior checks can miss, including performance, accessibility, and code quality. The rows also include an agent setup and reasoning-effort choice, so we preserve the published variants and costs without adding the scores to weighted coding rankings.
Snapshot
ReactBench score (Pass@1) on ReactBench — September 15, 2026 snapshot
We mirror the published reactbench score (pass@1) view for ReactBench. GPT-5.6 Sol leads the public snapshot at 46.7%, followed by Fable 5.1 · Max (45.6%) and Fable 5.1 · XHigh (44.1%). We do not use these results to rank models overall.
GPT-5.6 Sol
OpenAI
Fable 5.1 · Max
Anthropic
Fable 5.1 · XHigh
Anthropic
54 modelsCodingCurrentDisplay onlyUpdated September 15, 2026 snapshot
ReactBench score (Pass@1) table (54 models)
ScoreThe published ReactBench snapshot places GPT-5.6 Sol first at 46.7%. The third row is 2.6 points behind. The broader top-10 range is 6.2 points, so many of the published results sit in a relatively narrow band.
54 models have been evaluated on ReactBench. The benchmark falls in the Coding category. This category carries a 20% weight in BenchLM.ai's overall scoring system. ReactBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About ReactBench
Year
2026
Tasks
51 production React tasks
Format
Pass@1 weighted rubric score
Difficulty
Production frontend engineering
ReactBench uses 51 tasks drawn from real React repositories. Its weighted rubric score assigns zero to solutions that miss a blocking criterion. We mirror each published reasoning-effort variant and its average rollout cost, but keep the results out of weighted coding scores because the agent setup and effort level are part of the row.
BenchLM freshness & provenance
Version
ReactBench 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
FAQ
What does ReactBench measure?
A coding-agent benchmark for realistic React work, with rubrics that check production concerns such as performance, accessibility, correctness, and code quality.
Which model leads the published ReactBench snapshot?
GPT-5.6 Sol currently leads the published ReactBench snapshot with 46.7% reactbench score (pass@1). BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on ReactBench?
The September 15, 2026 snapshot snapshot contains 54 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 2,000+ readers.
One email each week. Unsubscribe anytime.