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.

← Labs

Microsoft

11 models tracked · known for Phi-3.5-mini-instruct, Phi-tiny-MoE-instruct, phi-4

11
Models tracked
11
In our data
88.8
Best score
3
Notable

Models

phi-3-medium 14Bindependently reproduced88.8phi-3-small 7.4Bself-reported87.6phi-3-mini 3.8Bself-reported84Phi-4independently reproduced79.73Phi-2self-reported67.87Phi-1.5self-reported46.8Phi-3.5-mini-instructunverified—Phi-tiny-MoE-instructunverified—phi-4unverified—phi-2unverified—Phi-3-mini-4k-instructunverified—
50 source records · huggingface_models, epoch_benchmarks. Every score is evidence-linked.

Follow the record

Other labs

  • Qwen
  • Meta AI
  • OpenAI
  • DeepSeek
  • NVIDIA
  • Google DeepMind
  • EleutherAI
  • Zhipu AI

Explore the record

  • Benchmark Integrity Index
  • Model Claim Ledger
  • Head-to-head comparisons