GPT-5.6 Luna vs Gemini 3.7 Flash: the sub-$1 tier
These are the two cheapest mainstream models in the index. Both are priced under a dollar per million input tokens, which puts high-volume workloads — batch extraction, classification, log triage — within reach of small budgets. This pair is about the floor of the market: what it costs to run a genuinely high-volume job without a frontier model budget.
GPT-5.6 Luna vs Gemini 3.7 Flash: the numbers side by side
| GPT-5.6 Luna | Gemini 3.7 Flash | |
|---|---|---|
| Gateway rate — input | $0.1 | $0.375 |
| Gateway rate — output | $0.6 | $1.875 |
| Official list — input | $0.2 | — |
| Official list — output | $1.2 | — |
| Blended at 1/3 output | $0.2667 | $0.875 |
| Output : input price ratio | 6.0× | 5.0× |
USD per 1M tokens. Gateway rates checked 2026-09-16; official list prices as published by each vendor. Gateway rates move with upstream promotions — re-check before committing a budget.
What each costs per month
| Monthly tokens | GPT-5.6 Luna | Gemini 3.7 Flash | Cheaper |
|---|---|---|---|
| 10M | $2.67 | $8.75 | GPT-5.6 Luna — 70% less |
| 100M | $26.67 | $87.5 | GPT-5.6 Luna — 70% less |
| 1000M | $267 | $875 | GPT-5.6 Luna — 70% less |
Assumes 1/3 of tokens are output and 2/3 input — roughly a chat or coding-agent mix. At a 100M tokens/month workload the gap is $60.83 a month, or $730 a year. Change the split in the calculator — output-heavy workloads widen the gap.
How much tokens each budget buys
| Budget | GPT-5.6 Luna | Gemini 3.7 Flash | Extra tokens |
|---|---|---|---|
| $100 | 375M | 114M | GPT-5.6 Luna — 261M more |
| $1,000 | 3,750M | 1,143M | GPT-5.6 Luna — 2,607M more |
Tokens purchasable at the blended rate above. Output tokens cost more than input, so reading these figures as "all input" would overstate them.
Which one to pick
When input pricing drops below a dollar per million tokens, the constraint moves from budget to throughput and accuracy. Practical considerations:
- Volumes here are large by definition. A workload that made sense on a frontier model budget is usually a much larger workload at this tier — re-run the throughput numbers, not just the cost numbers.
- Batch instead of streaming where you can. Long-running batch jobs smooth load and suit cheap models well; interactive latency requirements push you back toward dearer options.
- Watch for tiering. Some vendors charge more above a context threshold, so a large context window is not automatically priced at the headline rate. Check the model page for notes.
Switching between them
Both models are called through the same endpoint, so this is a model-ID change, not a migration. Setup references: OpenAI & Anthropic SDK · Claude Code · Cursor · reducing API cost.
Related comparisons: Claude vs Gemini · Claude vs GPT · Lowest-cost LLM APIs
FAQ
Is GPT-5.6 Luna or Gemini 3.7 Flash cheaper?
At current gateway rates GPT-5.6 Luna is cheaper for a mixed workload — $0.2667 versus $0.875 per 1M tokens blended at one third output. The answer depends on your input/output split: output tokens are priced several times higher than input on both models (6x and 5x respectively), so an output-heavy workload widens the gap.
How much does GPT-5.6 Luna cost per month?
At $0.2667 per 1M blended tokens: about $2.67 for 10M tokens a month, $26.67 for 100M, and $267 for 1B. That is pure token spend — there is no subscription on either model.
How much does Gemini 3.7 Flash cost per month?
At $0.875 per 1M blended tokens: about $8.75 for 10M tokens a month, $87.5 for 100M, and $875 for 1B.
Can I switch between them without rewriting my code?
Both are reached through the same gateway endpoint, so switching is a model-ID change in the request rather than a new integration. See the SDK guide for the one-line change, or the Claude Code setup if you are switching inside a coding agent.
What is the cheapest way to run a large batch job?
Price the job at the blended rate, not the input rate, and include retries. If the job is input-heavy and mechanical, the cheapest tier here is usually enough; if it needs verification, cost the verification pass too, because a cheap model plus a second opinion can still beat a dear model used once.