jevbench.dev · v1 preview
Leaderboard
Measured results and official citations. Jev models, gaming harnesses, and decision-model benchmarks. No fabricated win rates.
Models tested
8
Harness runs
38
Games tested
1
StarCraft II
Wins verified
0
Honest results
Jev-class decision models against a shared test suite
| # | Model | Use case | Category | Score | Speed | Cost in | Cost out | Notes |
|---|
Rows marked OURS are measured here; OFFICIAL cites upstream cards/benches.
Speed & Cost: Speed values are vendor claims unless marked "measured". FREE means no model API fee (hardware and upstream LLM costs may still apply for local/self-hosted models). Sources: TypeSafe blog, Gemini pricing, Laya-MLX BENCHMARKS.
Gaming harness results by game and provider — StarCraft II first
| # | Game | Agent / model | Provider | Wins | Runs | Notes |
|---|
Early harness results — no wins yet on MarineMicro unwinnable-as-Terran scenario. Campaign pivots incoming.
Harness v0.1 · 38 SC2 runs · 1 game (StarCraft II) · updated 20 Sep 2026