GPT-5 Nano vs Llama 4 Maverick 17 B128 E Instruct Fp 8: cost comparison
Prices updated 2026-11-17 · computed from live provider data
Short answer: Llama 4 Maverick 17 B128 E Instruct Fp 8 is cheaper on a
3:1 input/output blend — $0.062 per million tokens versus
$0.138, a 55% difference.
GPT-5 Nano offers the larger context window at 272,000 tokens, against 131,072 for Llama 4 Maverick 17 B128 E Instruct Fp 8.
Head to head
| GPT-5 Nano | Llama 4 Maverick 17 B128 E Instruct Fp 8 | |
|---|---|---|
| Input, per 1M tokens | $0.050 | $0.050 |
| Output, per 1M tokens | $0.400 | $0.100 |
| Blended 3:1, per 1M | $0.138 | $0.062 |
| Context window | 272,000 | 131,072 |
| Vendor | gpt | lambda_ai |
Monthly cost at real volumes
| Workload (input / output) | GPT-5 Nano | Llama 4 Maverick 17 B128 E Instruct Fp 8 |
|---|---|---|
| 1M in / 250K out | $0.150 | $0.075 |
| 10M in / 2.5M out | $1.50 | $0.750 |
| 100M in / 25M out | $15.00 | $7.50 |
Which should you pick?
Pick GPT-5 Nano when you need the longer context window (272,000 vs 131,072 tokens). Pick Llama 4 Maverick 17 B128 E Instruct Fp 8 when price per token dominates your workload — high-volume classification, extraction, or batch summarisation.