Skip to content
Mock data. Run values, controls, targets and reproductions on this preview are invented for layout. Claim-card text, upstream figures, the reference price and the RFC list come from the repository.
Inference ProofsMenu

Evidence, RFC-0009 §Targets

Targets

Relying parties publish thresholds they care about, attributed to them. Each target is evaluated against the least favourable completed repetition, not the mean. The project publishes no targets of its own; the two below are mock examples.

verify-under-60s

Example relying party (mock)

verify_seconds <= 60 s, scope verify, for claim class */exact-match/*

A mock target from the RFC-0009 example: a relying party that checks every proof on an 8-CPU verifier wants each check under a minute.

Evaluation of verify-under-60s
ConfigurationResultMeanLeast favourable repetitionLabelsReason
flashprove fp-maverick-w1024-b1meets3.113 s3.140 sscope verifygrade measuredtier reportedmock
flashprove fp-maverick-w1024-b24meets3.120 s3.150 sscope verifygrade measuredtier reproducedmock

cost-under-0.05-per-1k

Example relying party (mock)

cost_per_1k_tokens_whole_wall <= 0.05 USD/1k_token, scope whole_wall, for claim class *

A mock target that matches every claim class, so every prover configuration is evaluated against it.

Evaluation of cost-under-0.05-per-1k
ConfigurationResultMeanLeast favourable repetitionLabelsReason
flashprove fp-maverick-w1024-b1does_not_meet0.6301 USD/1k_token0.6320 USD/1k_tokenscope whole_wallgrade derivedtier reportedmock
flashprove fp-maverick-w1024-b24meets0.03487 USD/1k_token0.03493 USD/1k_tokenscope whole_wallgrade derivedtier reproducedmock
sash sash-l31-8b-c1024does_not_meet1.514 USD/1k_token1.528 USD/1k_tokenscope whole_wallgrade derivedtier reportedmock
sash sash-l31-8b-c4096does_not_meet1.417 USD/1k_token1.434 USD/1k_tokenscope whole_wallgrade derivedtier peer-reviewedmock