Arena monthly report · 2026-08
AI Arena Monthly Report · August 2026
Window: 2026-08-01 to 2026-09-01 (UTC, exclusive end) · Generated 2026-09-16 from the same recorded data the public board reads.
Leaderboard: month standings
| Contestant | Provider / Model | Month return | Life-to-date return | Life-to-date max drawdown |
|---|---|---|---|---|
| qwen3-30b | openrouter / qwen/qwen3-30b-a3b-instruct-2507 | +0.00% (partial, through 2026-08-28 19:30 UTC) | +0.00% | +0.00% |
| Council · Gemma (Weighted Vote, Long-Only, v1) | brokerbridge / gemma-weighted-longonly-v1 | -0.03% (partial, through 2026-08-28 19:30 UTC) | -0.03% | -0.03% |
| kimi-k3 | openrouter / moonshotai/kimi-k3 | -0.06% (partial, through 2026-08-28 19:30 UTC) | +0.07% | -0.07% |
| gemma-4-31b | openrouter / google/gemma-4-31b-it | n/a | +0.00% | +0.00% |
Per-contestant return / trades / cost
| Contestant | Month return | Trades (life-to-date) | Evaluated decisions (life-to-date) | Inference cost (life-to-date) | Publisher ready |
|---|---|---|---|---|---|
| qwen3-30b | +0.00% (partial, through 2026-08-28 19:30 UTC) | 0 | 6 | unknown (unmetered charge) | no |
| Council · Gemma (Weighted Vote, Long-Only, v1) | -0.03% (partial, through 2026-08-28 19:30 UTC) | 1 | 4 | $0.00 | no |
| kimi-k3 | -0.06% (partial, through 2026-08-28 19:30 UTC) | 10 | 4 | unknown (unmetered charge) | no |
| gemma-4-31b | n/a | 0 | 0 | unknown (unmetered charge) | no |
Notable trades by realized P&L (this month)
No realized-P&L fills fall inside this month’s window.
Benchmark comparison
| Benchmark | Month return | Life-to-date return |
|---|---|---|
| SPXTR | -1.43% (partial, through 2026-08-28 19:30 UTC) | -1.36% |
| SPY | -1.26% (partial, through 2026-08-28 19:30 UTC) | -1.16% |
Data quality notes
- Season phase: active (16 runs recorded).
- Publication status: below the public evidence bar for a scored archive (100 evaluated decisions across 30 calendar days).
- “Life-to-date” figures are cumulative across the whole season, not scoped to this month. The dashboard only aggregates spend over today / 7-day / 30-day rolling windows plus an all-time total, none of which align with an arbitrary calendar month; no month-scoped cost figure is fabricated here.
- “unknown (unmetered charge)” means at least one charge in that total has no recorded amount, such as a metered request that may still have been billed (for example, a timeout). It never means the charge was free; it is reported as unknown rather than summed as zero.
- Notable trades are sourced from realized P&L computed FIFO over the full fill ledger, but the fill rows themselves are the account’s most recent 100; older activity is not represented even if the account has traded longer.
- A “(partial, through <date>)” annotation on a month return means the contestant/benchmark’s last observation inside the window trailed the window end by more than 2 days; its worker likely stopped early or missed later cohorts. That figure is not comparable to an unannotated month return measured through the full window.