Live from GET /api/gspc — recompute anything, free
The GSPC board
22 axes · 15 measured · deterministic grading on frozen, published splits · a TIE means the leader's edge is statistically indistinguishable (McNemar p≥0.05) — ties are never counted as wins. Empty cells stay empty.
22 axes are on the board; 15 of them carry a measurement and 7 are declared slots with no run behind them. The larger number counts slots, the smaller counts measurements — quote both or quote the smaller. A published slot exists so the gap is visible; it is not evidence of anything having been measured.
Axis deep-dive
swarm
Bench SwarmBench v2b · n=37 · leader accuracy ≥38.4% · separation SEPARATED
Frozen gold bank (Hugging Face)Verify the signed chainRaw JSON (GET /api/gspc)Full board
| Axis | Bench | n | Leader accuracy | 95% CI | Separation |
|---|---|---|---|---|---|
| governance | GovBench | 237 | 70.0% | 63.9–75.5% | SEPARATEDp=0.0086 |
| safety | DefBench | 36 | 94.4% | 81.9–98.5% | TIE — indistinguishablep=0.6875 |
| provenance | ProvBench | 32 | 78.1% | 61.2–89.0% | TIE — indistinguishablep=0.7744 |
| continuity | PQCBench | 33 | 60.6% | 43.7–75.3% | TIE — indistinguishablep=1 |
| conformance | MCPBench | 35 | 74.3% | 57.9–85.8% | TIE — indistinguishablep=1 |
| openness | OSSBench | 32 | 87.5% | 71.9–95.0% | TIE — indistinguishablep=1 |
| machinery-conformity | MachBench | 33 | 54.5% | 38.0–70.2% | TIE — indistinguishablep=0.5811 |
| care | CareBench | 199 | 53.5% | 46.6–60.3% | SEPARATEDp=0.0356 |
| cross-reality | XRAIV | 32 | 81.2% | 64.7–91.1% | TIE — indistinguishablep=0.0654 |
| detector-interop | DetBench | 33 | 87.9% | 72.7–95.2% | TIE — indistinguishablep=0.4531 |
| art5-safeguard | Art5Bench | 36 | 97.2% | 85.8–99.5% | TIE — indistinguishablep=1 |
| swarm | SwarmBench v2b | 37 | ≥38.4%lower bound | withheld (n not independent) | SEPARATED |
| affect | AffectBench | 41 | 87.8% | 74.5–94.7% | SEPARATEDp=0.0078 |
| jail | GoldBank-Detector | 71 | 59.2% | 47.5–69.8% | TIE — indistinguishable |
| provenance-controls | ChainFacts | 6issuer accounts (not bank items) | no leader accuracy6 of the 16 instruments named in the registry | not applicable — no accuracy to bound | MEASURED — deterministic factsnot applicable — no fleet, no leader |
| reserve-attestation | — | 0nothing measured | unmeasured | unmeasured | UNMEASURED — not yet gated |
| regulatory-framework | — | 0nothing measured | unmeasured | unmeasured | UNMEASURED — not yet gated |
| distribution-integrity | — | 0nothing measured | unmeasured | unmeasured | UNMEASURED — not yet gated |
| custody-disclosure | — | 0nothing measured | unmeasured | unmeasured | UNMEASURED — not yet gated |
| ai-economy-index | — | 0nothing measured — 2 of 4 input components exist, no index computed | unmeasured | unmeasured | UNMEASURED — not yet gated |
| human-labour-index | — | 0nothing measured — 2 of 4 input components exist, no index computed | unmeasured | unmeasured | UNMEASURED — not yet gated |
| humanoid-labour-index | — | 0nothing measured — no input bank exists at all | unmeasured | unmeasured | UNMEASURED — not yet gated |
Arena Elo — signed
Per-axis winner-specific Elo from the live arena (12 models, 17 arena axes — the arena's own set, not the board's count above). Every score carries n + 95% CI. This leaderboard is signed — verify it below.
| Model | Elo | Games | Win-rate | 95% CI |
|---|---|---|---|---|
| mistral:7b | 1624.7 | 797 | 0.723 | 0.691–0.753 |
| qwen2.5:7b | 1578.1 | 899 | 0.689 | 0.658–0.718 |
| qwen3:4b | 1560.6 | 847 | 0.721 | 0.69–0.751 |
| qwen2.5:1.5b | 1545.4 | 887 | 0.417 | 0.385–0.45 |
| llama3:8b | 1518.8 | 28 | 0.536 | 0.358–0.705 |
| council-safe:latest | 1496.4 | 691 | 0.279 | 0.247–0.314 |
| qwen3:8b | 1481.3 | 15 | 0.4 | 0.198–0.643 |
| deepseek-r1:7b | 1448.8 | 24 | 0.458 | 0.279–0.649 |
| qwen2.5:0.5b-instruct | 1443.4 | 905 | 0.355 | 0.324–0.386 |
| council-oowm:latest | 1440.9 | 654 | 0.272 | 0.239–0.308 |
| qwen2.5-0.5b-cards:latest | 1439.2 | 160 | 0.475 | 0.399–0.552 |
| muse-glimmer:latest | 1422.3 | 197 | 0.386 | 0.321–0.455 |
In-lane — not board rows
Published as measured_in_lane on GET /api/gspc. Not counted in totals.public_count.
- slot15MEASURED
Slot15-Honesty
33 · n=35
- human-vs-aiMEASURED
Colosseum-Pairs
100 · n=35
Measurement, not certification. Leaders shown are point estimates (swarm quotes its 95% lower bound); only SEPARATED leads are statistically real — the live count is totals.separated_leads on GET /api/gspc. Jail is a measured floor when the stamp publishes one, never a hidden score. Full per-axis notes, fleet means and harm tails: GET /api/gspc. The living stamp carried there is marked UNVERIFIABLE — it does not reproduce under any published rule and is not a checkable attestation; the attestation over that payload that does verify is site_attestation.