ENPT

Crivo-150M

Last updated:

Open20–63ms$0/M input

Quadrant scores

See the full quadrant

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

  • Maturity4.5/10

    Public weights under CC BY-NC 4.0 (no commercial use), a thorough card with latency and contamination checks, local only. No hosted API, integrations or SLA.

  • Capability6.6/10

    Official boards (17 Oct): Decision Index 0.3.2 Full 22.21 (#134). Author latency 20 ms median on an A100, 63 ms on an M2 CPU. Context up to 32,768.

  • Adoption1/100

    ~94 downloads and 8 likes in three days. No integrations found.

Vendor claims

publicado (OK do Fabricio 17/10 BRT). Board audit of 17 Oct 2026: see the “Board audit” section below; where it conflicts with older board statements on this page, it wins.

Licence: CC BY-NC 4.0. You can’t use the weights commercially.

Crivo-150M is a small encoder (149M parameters, from cross-encoder/ettin-reranker-150m-v1) that answers typed questions (choice, noul, score) with calibrated probabilities. It is on the official Decision Index 0.3 board at 22.21 full, rank 120 of 160, which the author shows is the best entry at or below 150M (the Pareto-frontier claim is the author’s framing, UnverifiedClaim). A smaller sibling, Crivo-32M, sits at 11.87. Trained up to 8,192 tokens, it reads up to 32,768 without truncation. Use it where cost and a laptop CPU matter more than accuracy.

Score working (13 Oct 2026)

  • Maturity = 6 × 0.30 + 8 × 0.25 + 2 × 0.25 + 1 × 0.20 = 1.80 + 2.00 + 0.50 + 0.20 = 4.50 → 4.5
  • Capability = 9 × 0.30 + 6 × 0.25 + 7 × 0.25 + 5 × 0.20 = 2.70 + 1.50 + 1.75 + 1.00 = 6.95 → 7.0
  • Adoption = 2 × 0.35 + 1 × 0.35 + 0 × 0.30 = 0.70 + 0.35 + 0 = 1.05 → 1

Accuracy 5, not 8: the board result is verifiable, but 22.21 is low in absolute terms; we score the level, not only that it was verified. Availability 6 for the NC licence.

Board audit (17 Oct 2026)

publicado (OK do Fabricio 17/10 BRT). Checked against the official boards on 17 Oct 2026: Decision Index 0.3.2 (data of 16 Oct, 19:00 UTC), JevBench and ImageJevBench live. The boards are separate sources and are never averaged. Older board statements and rankings further down this page are superseded where they conflict with this section.

Board Row Result
Decision Index 0.3.2 (16 Oct 19:00 UTC) Crivo-150M Full 22.21 (#134, tie group 38); public 36.39 (#112); same-skills 27.74; new-domains 9.77
Decision Index 0.3.2 (16 Oct 19:00 UTC) Crivo-32M Full 11.87 (#151, tie group 45); public 25.06 (#140); same-skills 18.84; new-domains 0.92

Score update (audit, 17 Oct 2026). Capability 7.0 → 6.6 = types × 0.30 + calibration × 0.25 + latency × 0.25 + accuracy × 0.20 = 9 × 0.30 + 6 × 0.25 + 7 × 0.25 + 3 × 0.20 = 6.55 → 6.6. Types, calibration and latency come from the earlier working (9 / 6 / 7); only accuracy is re-derived from the board (3; was 5): Decision Index Full 22.21 ÷ 60.11 × 8 = 3.0, half-up. Maturity and adoption unchanged.

Hugging Face on 17 Oct (tardellirs/crivo-150m): 216 downloads, 11 likes, licence tag cc-by-nc-4.0, last modified 2026-10-12.