tensor.news
Integrity IndexModelsCompareLabsPapersMethodologyAbout
tensor.news

© 2026

Evidence

  • Integrity Index
  • Model Claim Ledger
  • Compare

Entities

  • Models
  • Benchmarks
  • Labs
  • Papers

Method

  • Methodology
  • About
  • Privacy

Machine

  • Feed (Atom)
  • llms.txt
  • llms-full.txt
  • Sitemap

tensor.news grades benchmarks and tags model claims by evidence status. Every fact carries a source record and provenance. Scores are framed as task performance under a disclosed harness, never as deployed capability.

← Models

PowerMoE-3b

IBM Research·released 2024-08-14

unverified

Scored on 11 benchmarks, leading 9. 0% of its scores are independently reproduced — the rest are self-reported or unverified.

11
Benchmarks scored
9
SOTA leads
79.1
Best score
0%
Reproduced
0 of 11

Leads on: ARC, BoolQ, GSM8k (5 shot), Hellaswag, MBPP, MMLU (5 shot), PIQA, humaneval, math (4 shot)

Benchmark fingerprint

Every score, ranked, colored by evidence status.

  • PIQA
    79.1
  • Hellaswag
    71.5
  • BoolQ
    65
  • Winogrande
    65
  • ARC
    58.1
  • MMLU (5 shot)
    42.8
  • OpenBookQA
    41
  • MBPP
    32.4
  • GSM8k (5 shot)
    25.9
  • humaneval
    20.1
  • math (4 shot)
    14.8

All benchmark scores

BenchmarkScoreEvidence statusMeasured
PIQA79.1accuracy-normunverified—
Hellaswag71.5accuracy-normunverified—
BoolQ65accuracyunverified—
Winogrande65accuracy-normunverified—
ARC58.1accuracy-normunverified—
MMLU (5 shot)42.8accuracyunverified—
OpenBookQA41accuracy-normunverified—
MBPP32.4pass@1unverified—
GSM8k (5 shot)25.9accuracyunverified—
humaneval20.1pass@1unverified—
math (4 shot)14.8accuracyunverified—
1 source records · huggingface_models. Every score is evidence-linked.

Follow the record

More from IBM Research

  • granite-4.1-8b

Head-to-head

  • vs Claude Fable 5
  • vs GPT-5.5
  • vs Gemini 3.1 Pro
  • vs Llama 3.1-405B
  • vs phi-3-small 7.4B
  • vs DeepSeek-V3

Benchmarks it leads

  • PIQA