Versioned benchmark collection

Space Flight v2

Harness v2 · 4 measured runs across 1 target. Each entry is a single recorded attempt; repeat-run reliability is not established.

Inspect the run ledgers ← All collections

Astronaut approaching a landed spacecraft in the Space Flight v2 collection

4Measured runs

1Frozen targets

4Model families

4Configurations

Collection scope

Recorded evidence, without a synthetic score.

Runs share the same frozen target and prompt within each test, with disclosed provider-specific stacks. Costs are recorded estimates as labelled in each ledger, not subscription invoices.

Targets

Space Flight Stage 2 — Landfall

Model coverage

Claude Opus 5 · Max
Claude Fable 5 · Max
GPT-5.6 Sol · Ultra
Kimi K3 · Max

Evidence available

Four measured run ledgers
Four public presentation recordings
Cost, token, workflow timing, and artifact receipts

Still missing

Repeat-run reliability evidence
Standardized gameplay and physical-input verification
Comparable local FPS measurements

Public receipts

4 measured run ledgers

Open any row for its recorded workflow cost, timing, token usage, stack disclosure, artifact integrity, and explicit evidence gaps.

01Claude Opus 5Max reasoning · Space Flight Stage 2 — Landfall$133.83 estimated API-equivalent costrecorded estimate2h 31m 33.5s wall-clockEnd-to-end workflowOpen 02Claude Fable 5Max reasoning · Space Flight Stage 2 — Landfall$280.14 estimated API-equivalent costrecorded estimate2h 05m 20.4s wall-clockEnd-to-end workflowOpen 03GPT-5.6 SolUltra reasoning · Space Flight Stage 2 — Landfall$20.84 estimated API-equivalent costrecorded estimate1h 29m 57.7s wall-clockEnd-to-end workflowOpen 04Kimi K3Max reasoning · Space Flight Stage 2 — Landfall$15.58 estimated API-equivalent costrecorded estimate2h 23m 18.1s wall-clockEnd-to-end workflowOpen