whatmodel.sh / decision / TinyJev-4B (JevBench mini)
TinyJev-4B (JevBench mini)
↗ share
+ contribute
TinyJev-4B (JevBench mini) is a source-available decision implementation from AnkitAI.
Catalogue identity notes
JevBench 0.1.0 mini tested configuration tinyjev-4b: Decoder, pointer head; base Qwen3-4B-Base; readout merged LoRA + pointer head (Kev recipe). Separate identity because the report does not establish equivalence to the newer Decision Index checkpoint or serving recipe. No latency imported.
$ how it compares
The decision models ranked either side of TinyJev-4B (JevBench mini), out of 357 with a score.
| # | model | score | where |
|---|---|---|---|
| 211 | JevK5 v0.3 (JevBench mini) | 50±2 | source code |
| 212 |  FlyMy.AI Decision 4B v1.6 | 50±1 | model files |
| 213 | TinyJev-4B (JevBench mini) | 50±2 | source code |
| 214 |  Nari Decision 4B v3.2 | 49±1 | model files |
| 215 |  Hopper (G) 1.2 Qwen3.5-4B LoRA | 49±1 | model files |
see all decision models · find the ones that fit your machine
$ what it's good at
consistency▃·▂▂·▅▃·▃▅▃█ #9 of 21 · 65% of its score · 5 tests
Beats JevK5 v0.3 (JevBench mini) here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| JevBench mini 0.1.0 · Probability measure coherence (pass rate) | 42.2% ✎ | ▄·▁·▃▃▁·▃▄▁█ beats 11/20 | 48% |
| JevBench mini 0.1.0 · Representation consistency (pass rate) | 78.3% ✎ | ▃·▂·▂▂▂▃▅▅▃█ beats 11/20 | 14% |
| JevBench mini 0.1.0 · Logical coherence (pass rate) | 75.8% ✎ | ▃·▃·█▅▅██▃█▅ beats 9/20 | 2.3% |
| JevBench mini 0.1.0 · Choice-set coherence (pass rate) | 90.3% ✎ | ▁▃▄█▄▄·····▄ beats 17/20 | 1.2% |
| JevBench mini 0.1.0 · Batch independence (pass rate) | 100.0% ✎ | ▁··········█ beats 15/20 | 0.0% |
other tests1 not in a task · 33% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| JevBench mini 0.1.0 · unchanged-answer accuracy | 86.6% ✎ | ▁·▂·▂···▂▂▃█ beats 10/20 | 33% |
$ where to run it
Implementation and setup instructions ↗
Published implementation and setup documentation verified. This entry describes a decision readout or evaluated configuration, not necessarily separate trained weights. Follow the referenced version and model settings; hardware compatibility has not been verified. Source · checked 2026-10-31.
Model repository and documentation: github.com/ankit-aglawe ↗
$ how the score was worked out · JevBench 0.1.0 mini evaluatio… 98%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 JevBench 0.1.0 mini evaluation reports | 6 | 98% |
| prior · pull toward the middle | 2% | |
| weighted average | = 49.5 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json