← Forge leaderboard

Forge 2.0 · Independent builds

Flint 40B-A2.5B 4-bit

RecursiveSI/flint-int4

Median score
0.1123
Observed range
0.1073–0.1172
Independent builds
2

The median summarizes these 2 independent model builds; the range shows the lowest and highest observed scores. These builds share the same task and captured comparison cohort. Each build retains its own score, checks and evidence. The observed range is not a confidence interval.

Individual builds

Flint 40B-A2.5B 4-bit · local Mac Goose

0.1073
Model build
65m 53s
Started
Open Build 1 checks and evidence →

Flint 40B-A2.5B 4-bit · local Mac Goose

0.1172
Model build
94m 46s
Started
Open Build 2 checks and evidence →

Tier comparison

Each column shows one build's own tier scores. The overall median is calculated from the overall build scores.

TierBuild 1Build 2
L · Lint and bundles1.00000.9722
K · Platform currency0.97500.8650
T · Event pipeline0.91890.9189
R · Reconcile0.98180.9818
S · Storage0.75001.0000
B · Resolvers and permissions0.78700.7870
U · UI function0.75190.6649
V · Visual1.00001.0000
A · Rovo0.50000.5000
E · Excellence0.81600.7617
R1 · Migration0.75000.7500
R2 · Dosing0.28210.4349
R3 · Time limits0.93790.9379
R4 · Changing world0.16060.3455
R5 · Admin panel0.00000.0000
R6 · CI web trigger0.10000.1000
R7 · Custom field0.00000.0000
R8 · Forge LLM0.50000.0000
R9 · Boot budget0.44450.4445
R10 · Native JQL0.33330.5648
Comparison receipts
Task fingerprint
ef9776e6b12a87482319f25954a9d4969ed78b28d3b9ec7398933bb84b0da292
Captured comparison cohort
7b62513536cffc3e27ebc29e089492d1ca794a252766e7c5df6ae8aceca11443

How Forge 2.0 is scored