GPT-5.6 Luna vs MiniMax-M3: ultra-cheap utility LLM showdown
Luna is the absolute floor on per-token price for utility work. MiniMax-M3 earns its keep only when you need the published batch API for half-off workloads, or when MiniMax's specific reasoning profile suits the task better than OpenAI's.
Two utility-tier models priced like a rounding error on a frontier bill.
GPT-5.6 Luna and MiniMax-M3 are both in the ultra-cheap utility tier - designed for high-volume classification, routing, and formatting work where per-token cost dominates. Luna is the cheaper of the two on list price, but MiniMax-M3 has a real 1M context window and a published 50% batch discount.
By TechCompare · Updated
Cost Comparison
Based on 100,000 input tokens (50% cached), 5,000 output tokens, and 100 requests.Prices are fetched live from OpenRouter and may include temporary promotional discounts not accounted for in our article and comparison figures.
Side-by-side specs
| Spec | GPT-5.6 Luna | MiniMax-M3 |
|---|---|---|
| Input Cost (per M) | $0.20 (better on this spec) | $0.30 |
| Output Cost (per M) | $1.20 | $1.20 |
| Cached Input (per M) | $0.02 (better on this spec) | $0.06 |
| Batch Discount | No | 50% (better on this spec) |
| Context Window | 1.05M | 1M |
How they differ
GPT-5.6 Luna is priced at $0.20 per million input tokens and $1.20 per million output tokens at OpenAI's list price, with a 90% caching discount ($0.02 per million) and a 1.05M context. MiniMax-M3 is priced at $0.30 per million input tokens and $1.20 per million output tokens, with an 80% caching discount ($0.06 per million), a 1M context, and a published 50% batch discount on OpenRouter. For a 30K input + 3K output request, Luna costs $0.0096 vs MiniMax's $0.0126 - Luna is still cheaper at this volume. MiniMax-M3's offsetting strengths are its published batch discount (Luna has no batch on OR) and the larger per-call context fit. Note: OpenRouter currently runs a limited-time 50% discount that can show Luna at $0.10/$0.60 in live calculators.
Verdict
Luna wins the rows that decide utility pricing: $0.20/M input against $0.30 and $0.02/M cached input against $0.06, both roughly 3x cheaper. Output ties at $1.20/M and context is near-tied (1.05M versus 1M). The single row MiniMax takes is the 50% batch discount against Luna's none, dropping its batch output to $0.60/M. The verdict splits cleanly by traffic shape: live one-shot loads go to Luna, async bulk jobs go to MiniMax.
Which should you pick?
Choose GPT-5.6 Luna
Maximum-volume utility work - classification, formatting, routing. Luna's $0.20/$1.20 list-price rates make it the cheapest per-token option in the 2026 frontier lineups, and temporary discounts can drop it further.
Choose MiniMax-M3
Production routes that benefit from the 50% batch discount, or workloads where MiniMax's reasoning profile specifically outperforms Luna at the task.