A display-only MiniMax provider benchmark for end-to-end coding-agent and product-building tasks.
BenchLM mirrors the published score view for VIBE V2. MiniMax M3 leads the public snapshot at 50.1%. BenchLM does not use these results to rank models overall.
Year
2026
Tasks
End-to-end coding-agent tasks
Format
Task success rate
Difficulty
Frontier coding-agent workflows
MiniMax reports VIBE V2 in the M3 comparison chart. BenchLM tracks it as a display-only provider row because it is not part of the weighted coding schema.
Version
VIBE V2 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.
A display-only MiniMax provider benchmark for end-to-end coding-agent and product-building tasks.
MiniMax M3 by MiniMax currently leads with a score of 50.1% on VIBE V2.
1 AI models have been evaluated on VIBE V2 on BenchLM.
For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.