Sieve 31B v3 enters the Decision Index at #2, just behind GLiDE 1.2
publicado (OK do Fabricio 28/10 BRT).
The Decision Index 0.4.1 added four open models on 28 Oct 2026 (06:03 BRT). The strongest is Sieve-31B-v3 from Sthanika AI, an Apache-2.0 fine-tune of Gemma 4 31B published on 27 Oct.
| Decision Index 0.4.1 (data 28 Oct) | Full | Rank |
|---|---|---|
| Fastino GLiDE 1.2 | 71.52 | #1 |
| Sieve 31B v3 | 71.10 | #2 |
| vLLM-SR d3-ultra | 70.18 | #3 |
| WeMM-JEV-27B (new) | 68.49 | #5 |
| Ainvest-Decision-27B v1.0 (new) | 68.23 | #7 |
| MISA AI 2.0 Decision Plus · oak 1.1 (new) | 67.62 | #11 |
| Jev 1.13 | 59.22 | #51 |
v3 is a big step from the Sieve 31B the board measured on 16 Oct (Full 63.81). The same batch also brought WeMM-JEV-27B, Ainvest-Decision-27B and oak 1.1, which this site has not reviewed yet.
Score. The Sieve page keeps capability 7.9: its accuracy was already at the cap of 8.