Directory · Model comparison
GPT-5.6 Luna vs Llama 3.1 70B Instruct
Side-by-side API economics and benchmarks. Per-1M input and output rates, context window, and verified metrics.
Tale of the tape
Head-to-head specs pulled from each model profile.
| Metric | GPT-5.6 Luna | Llama 3.1 70B Instruct |
|---|---|---|
| Input cost / 1M | $0.20 | $0.40 |
| Output cost / 1M | $1.20 | $0.40 |
| Context window | 1,050,000 tokens | 128,000 tokens |
| MMLU score | — | 83.6% |
Cost verdict
Enter monthly input and output volumes to see total cost savings.
GPT-5.6 Luna
- Input: 1M × $0.20 = $0.20
- Output: 1M × $1.20 = $1.20
- Total
- $1.40
Llama 3.1 70B Instruct
- Input: 1M × $0.40 = $0.40
- Output: 1M × $0.40 = $0.40
- Total
- $0.80
Total cost savings
$0.60
Llama 3.1 70B Instruct saves $0.60 per month vs the other model at this volume.
Model routing savings
Uses this page's verified rates. GPT-5.6 Luna is the premium route (higher combined per-1M cost); Llama 3.1 70B Instruct handles routine traffic.
All premium
GPT-5.6 Luna
$70.00
100% of input and output on the premium route
Routed stack
80% → Llama 3.1 70B Instruct
20% → GPT-5.6 Luna
$46.00
- Routine slice
- $32.00
- Premium slice
- $14.00
Monthly savings
$24.00
34.3% lower
Rates pulled from this page's model database entries (gpt-5-6-luna, llama-3-1-70b-instruct) — same per-1M input/output figures as Tale of the tape. Production routing needs quality gates and monitoring.
Related comparisons
More head-to-head model matchups.