Skip to main content

Benchmark profile

MLS-Bench Lite

A 30-task subset of MLS-Bench that evaluates whether AI systems can invent generalizable and scalable machine-learning methods.

Data verified

Benchmark score on MLS-Bench Lite — July 23, 2026

BenchLM mirrors the published score view for MLS-Bench Lite. Kimi K3 leads the public snapshot at 48.3% , followed by Kimi K2.7 Code (35.1%). BenchLM does not use these results to rank models overall.

2 modelsCodingCurrentDisplay onlyUpdated July 23, 2026

Benchmark score table (2 models)

Score
1
Kimi K3Moonshot AI · Closed
48.3%
2
Kimi K2.7 CodeMoonshot AI · Open weight
35.1%

About MLS-Bench Lite

Year

2026

Tasks

30 machine-learning research tasks

Format

Agentic ML task evaluation

Difficulty

ML research and systems engineering

Moonshot reports MLS-Bench Lite as a coding-agent result for Kimi K2.7 Code. BenchLM stores the provider-reported exact value separately from weighted coding benchmarks because the row is a newly reported benchmark variant with sparse public model coverage.

BenchLM freshness & provenance

Version

MLS-Bench Lite 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

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 MLS-Bench Lite measure?

A 30-task subset of MLS-Bench that evaluates whether AI systems can invent generalizable and scalable machine-learning methods.

Which model scores highest on MLS-Bench Lite?

Kimi K3 by Moonshot AI currently leads with a score of 48.3% on MLS-Bench Lite.

How many models are evaluated on MLS-Bench Lite?

2 AI models have been evaluated on MLS-Bench Lite on BenchLM.

Compare Top Models on MLS-Bench Lite

Last updated: July 23, 2026 · BenchLM version MLS-Bench Lite 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.