Head-to-head showdown: Llama 4 Scout (109B MoE) ($0.15 in / $0.45 out per 1M) vs Amazon Nova Premier ($2.00 in / $8.00 out per 1M). Llama 4 Scout (109B MoE) is 16.7× cheaper across standard token mixes, with 10M vs 1M context windows.
meta-llama/llama-4-scout · meta
amazon.nova-premier-v1:0 · amazon
Benchmark
| Workload Scenario | Llama 4 Scout (109B MoE) | Amazon Nova Premier | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $0.15 | $2.00 | −$1.85 |
| 1M output tokens (generation) | $0.45 | $8.00 | −$7.55 |
| 1M tokens · 70% input / 30% output mix | $0.24 | $3.80 | −$3.56 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.00096 | $0.0108 | −$0.00984 |
| Monthly scale (10K requests / day) | $288.00 | $3,240.00 | −$2,952.00 |
Negative difference = Llama 4 Scout (109B MoE) is cheaper. Positive = Amazon Nova Premier is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
Llama 4 Scout (109B MoE) is cheaper on both input and output rates, so it costs less at every input/output mix. Price alone still isn't the whole decision: capability, latency and context limits (Llama 4 Scout (109B MoE): 10M, Amazon Nova Premier: 1M) may justify the premium for your task.
FAQ
On input, Llama 4 Scout (109B MoE) is cheaper ($0.15/M vs $2.00/M). On output, Llama 4 Scout (109B MoE) is cheaper ($0.45/M vs $8.00/M). The same model is cheaper on both sides, so it wins at every mix.
A chat-style request (4,000 input + 800 output tokens, 50% cached) costs $0.00096 on Llama 4 Scout (109B MoE) and $0.0108 on Amazon Nova Premier — Llama 4 Scout (109B MoE) is 11.3× cheaper for that workload.
Llama 4 Scout (109B MoE) supports 10,000,000 tokens (16.4K max output); Amazon Nova Premier supports 1,000,000 (32.8K max output). Llama 4 Scout (109B MoE) fits 10.0× more context, which matters for long documents and agents.