live eventEidos is moving itself to 90%-cheaper silicon without losing intelligence. Watch.follow on LinkedIn →
mission: 90% local authorship — waiting for first eventhow?
← charts + scores

llama3.1:8b on A6000, A100, H100 — 2026-04-17

2026-04-17T21:56:19.000Z 2026-04-17T21:57:39.630Z
tweet →

On 2026-04-17, llama3.1:8b ran across A6000, A100, H100. H100 finished first at 191.4 tok/s, 26× faster than A6000. The per-million-token cost: $3.61 (H100) vs $13.32 (A6000) — the surprise: the headline GPU H100 was also the cheapest per-million tokens, coming in 3.7× less than A6000.

Podium

H100
1st
thunder-h100
peak tok/s
191.4
avg tok/s
191.4
$ / 1M tok
$3.61
A100
2nd
thunder-a100
peak tok/s
13.7
avg tok/s
13.7
$ / 1M tok
$15.82
A6000
3rd
thunder-a6000
peak tok/s
7.3
avg tok/s
7.3
$ / 1M tok
$13.32

pit wallH100 just posted 193 tok/s on llama3.1:8b — 13.7× faster than the A100 lane (1428h ago).

llama3.1:8b on A6000, A100, H100 — 2026-04-17 — Eidos Live · Crucible