Nova Lite vs Qwen 2.5 32 B Instruct: cost comparison
Prices updated 2026-10-30 · computed from live provider data
Short answer: Qwen 2.5 32 B Instruct is cheaper on a
3:1 input/output blend — $0.095 per million tokens versus
$0.105, a 10% difference.
Nova Lite offers the larger context window at 300,000 tokens, against 128,000 for Qwen 2.5 32 B Instruct.
Head to head
| Nova Lite | Qwen 2.5 32 B Instruct | |
|---|---|---|
| Input, per 1M tokens | $0.060 | $0.060 |
| Output, per 1M tokens | $0.240 | $0.200 |
| Blended 3:1, per 1M | $0.105 | $0.095 |
| Context window | 300,000 | 128,000 |
| Vendor | amazon | nebius |
Monthly cost at real volumes
| Workload (input / output) | Nova Lite | Qwen 2.5 32 B Instruct |
|---|---|---|
| 1M in / 250K out | $0.120 | $0.110 |
| 10M in / 2.5M out | $1.20 | $1.10 |
| 100M in / 25M out | $12.00 | $11.00 |
Which should you pick?
Pick Nova Lite when you need the longer context window (300,000 vs 128,000 tokens). Pick Qwen 2.5 32 B Instruct when price per token dominates your workload — high-volume classification, extraction, or batch summarisation.