ENPT

WeMM-JEV-27B

Last updated:

Open98–98ms$0/M input

Quadrant scores

See the full quadrant

Scored with the public quadrant rubric (maturity × capability; bubble = adoption). Revise as evidence lands.

  • Maturity4.8/10

    Apache-2.0 full fine-tune of Qwen3.8-27B (26.1B, 11 safetensors shards) with its own D3 engine, HTTP server and Decision Index adapter. No hosted API, integrations or SLA.

  • Capability7.9/10

    Decision Index 0.4.1 (29 Oct) Full 68.49 (#8 of 247), ECE 0.051, median 98 ms, p95 1.08 s. Choice, yes/no (noul) and score over text and images. No calibration method published.

  • Adoption1/100

    Rover00/WeMM-JEV: about 315 downloads and 1 like (31 Oct); a mirror (mwei/WeMM-JEV) appeared on 31 Oct. No integrations or production users found.

Vendor claims

publicado (curadoria autônoma 31/10 BRT). Scores are calculated from scratch.

What it is

WeMM-JEV is a 27B multimodal decision model fine-tuned from Qwen3.8-27B by the Hugging Face user Rover00. You send a state (text, plus images if you want) and typed questions; it returns a probability for every option in one forward pass per question. It ships its own inference package (D3Engine), an HTTP server and an adapter for the Decision Index kit. Source: Hugging Face · results dataset.

A second copy, mwei/WeMM-JEV, appeared on 31 Oct with the same card and file layout. We treat Rover00 as the source.

Board results

Board Row Result
Decision Index 0.4.1 (data 29 Oct) WeMM-JEV-27B Full 68.49 (#8 of 247), public 69.37; ECE 0.051; median 98 ms, p95 1.08 s
JevBench (v1.8.1, 30 Oct) WeMM-JEV Measured on the 27 Oct draw, shown unranked until that draw is equated

Not on the Artificial Analysis Decision Index (31 Oct). The boards are separate and are never combined.

Vendor claims (UnverifiedClaim)

  • The card reports public Decision Index 0.3 scores of 69.37 for text and 77.97 for vision. The text figure matches the board’s public split; the vision figure is the author’s own run.

Scores (from scratch, 31 Oct 2026)

  • Capability 7.9 = types × 0.30 + calibration × 0.25 + latency × 0.25 + accuracy × 0.20 = 10 × 0.30 + 6 × 0.25 + 7 × 0.25 + 8 × 0.20 = 3.00 + 1.50 + 1.75 + 1.60 = 7.85 → 7.9. Types 10: choice, yes/no and score. Calibration 6: board ECE 0.051, no published method. Latency 7: board median 98 ms, but p95 1.08 s (same step as other ~100 ms board medians). Accuracy: 8 × 68.49 ÷ 59.22 = 9.25 → capped at 8 (one board, not #1).
  • Maturity 4.8 = availability × 0.30 + docs × 0.25 + integrations × 0.25 + support × 0.20 = 7 × 0.30 + 7 × 0.25 + 2 × 0.25 + 2 × 0.20 = 2.10 + 1.75 + 0.50 + 0.40 = 4.75 → 4.8.
  • Adoption 1 = engagement × 0.35 + ecosystem × 0.35 + production × 0.30 = 1 × 0.35 + 1 × 0.35 + 0 × 0.30 = 0.70 → 1.