qwen2.5:1.5b on A6000, A100, H100 — 2026-04-18
2026-04-18T01:23:09.000Z → 2026-04-18T01:24:26.966Z
On 2026-04-18, qwen2.5:1.5b ran across A6000, A100, H100. H100 finished first at 407.8 tok/s, 14× faster than A6000. The per-million-token cost: $1.70 (H100) vs $3.33 (A6000) — the surprise: the headline GPU H100 was also the cheapest per-million tokens, coming in 2.0× less than A6000.
Podium
H100
1st
thunder-h100
- peak tok/s
- 407.8
- avg tok/s
- 407.8
- $ / 1M tok
- $1.70
A100
2nd
thunder-a100
- peak tok/s
- 43.9
- avg tok/s
- 43.9
- $ / 1M tok
- $4.94
A6000
3rd
thunder-a6000
- peak tok/s
- 29.2
- avg tok/s
- 29.2
- $ / 1M tok
- $3.33