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

qwen2.5:14b on A6000, A100, H100 — 2026-04-17

2026-04-17T22:31:15.000Z 2026-04-17T22:32:09.758Z
tweet →

On 2026-04-17, qwen2.5:14b ran across A6000, A100, H100. H100 finished first at 107.9 tok/s, 13× faster than A100. The per-million-token cost: $6.41 (H100) vs $27.08 (A100) — the surprise: the headline GPU H100 was also the cheapest per-million tokens, coming in 4.2× less than A100.

Podium

H100
1st
thunder-h100
peak tok/s
107.9
avg tok/s
107.9
$ / 1M tok
$6.41
A6000
2nd
thunder-a6000
peak tok/s
49.7
avg tok/s
49.7
$ / 1M tok
$1.96
A100
3rd
thunder-a100
peak tok/s
8.0
avg tok/s
8.0
$ / 1M tok
$27.08

tickerH100 just posted 193 tok/s on llama3.1:8b — 13.7× faster than the A100 lane (1428h ago).

qwen2.5:14b on A6000, A100, H100 — 2026-04-17 — Eidos Live · Crucible