Attested Run id run-90f62a5afcfb

claude-haiku-4-5

on gsm8k · anthropic-claude · n=8 ⚠ statistically thin · Sun, 26 Apr 2026 19:26:15 GMT
100.0
±16.2 · 95% CI [67.6, 100.0]
claude-haiku-4-5 scored 100.0 on gsm8k across 8 problems. The transcript is committed to Merkle root sha256:4d7cc7ff5b489d0… and signed by attestor benchlist-vercel-inline-0 with Ed25519 signature 03fee5c994a957a3cdcea5…. The signature is verified in your browser below — no server round-trip required.
Raw JSON ↗ Replay for $0.50
Dataset hashsha256:09a35a0a0a48f13840457c82e2c2da6a7884ec21b51154139867843c2e4da5c7
Methodology hashsha256:144e8efdcdb66a248c57935cea7c8d00cbc6c287341355ab753cc5f445238bfb
Merkle rootsha256:4d7cc7ff5b489d00aee230a9a89ab3d2aca91bd38b7e16f297e7e8a3d8c035d1
Attestor pubkeycb6e95d0f7b402e254f491b57767df3a3a93ae92f1faee3a02aa52e728f5cd11
Signature03fee5c994a957a3cdcea5195327666f03d4fef685bf1fb6d5ddcbf840db416e8d7d15c02e6b5256afc04b68fade73a6594b071e1e0699087868183d6d50dc06
Runnerbenchlist-vercel-inline@1.0.0
Started2026-04-26T19:26:02.768Z
Finished2026-04-26T19:26:15.228Z
Runner provenance · what code produced this
how to verify →
versionbenchlist-vercel-inline@1.0.0 commit3b3562b5a0ff421dfb1530aaef628667a31dbd93 repogithub.com/benchlist/runner adaptersha256:inline-js:gsm8k judgesha256:inline-js:scoreOne digestsha256:7e9f1dfe479c7831a6b11092144a02bab7c52861802c40dbca2faeaec47a46d1
Not yet anchored on-chain. Anyone can anchor for ~$0.01–$1.40 in gas → /anchor?run=run-90f62a5afcfb
Best per benchmark → gsm8k guide → Anchor on-chain → Dispute