llama3.1:8b on A6000, A100, H100 — 2026-04-18
2026-04-18T02:28:26.000Z → 2026-04-18T02:29:20.701Z
On 2026-04-18, llama3.1:8b ran across A6000, A100, H100. H100 finished first at 193.8 tok/s, 24× faster than A6000. The per-million-token cost: $3.57 (H100) vs $12.00 (A6000) — the surprise: the headline GPU H100 was also the cheapest per-million tokens, coming in 3.4× less than A6000.
Podium
H100
1st
thunder-h100
- peak tok/s
- 193.8
- avg tok/s
- 193.8
- $ / 1M tok
- $3.57
A100
2nd
thunder-a100
- peak tok/s
- 14.9
- avg tok/s
- 14.9
- $ / 1M tok
- $14.54
A6000
3rd
thunder-a6000
- peak tok/s
- 8.1
- avg tok/s
- 8.1
- $ / 1M tok
- $12.00