# MiniMax M2.7 vs MiniMax M3

As of October 21, 2026, MiniMax M3 costs 20% more per input token than MiniMax M2.7 ($0.30 vs $0.25 per 1M) and has a 1M-token context window versus 192K.

- Page: https://models.litellm.ai/compare/minimax-m2.7-vs-minimax-m3
- MiniMax M2.7: https://models.litellm.ai/models/minimax-m2.7
- MiniMax M3: https://models.litellm.ai/models/minimax-m3

## Side by side

| | MiniMax M2.7 | MiniMax M3 |
|---|---|---|
| Input, per 1M tokens | $0.25 | $0.30 |
| Output, per 1M tokens | $1.00 | $1.20 |
| Cache read, per 1M tokens | $0.05 | $0.06 |
| Cache write, per 1M tokens | – | – |
| Context window | 196,608 tokens | 1,000,000 tokens |
| Max output | 204,800 tokens | 128,000 tokens |
| Image input | No | Yes |
| Tool calling | Yes | Yes |
| Structured output (JSON) | Yes | Yes |
| Prompt caching | Yes | Yes |
| Reasoning | Yes | Yes |
| Providers on LiteLLM | 7 | 10 |
| Price shown is from | DeepInfra | MiniMax |
| Added to LiteLLM | May 2026 | June 2026 |

Headline prices in USD per 1M tokens, from LiteLLM's model price file.

## Cheapest provider for each

- MiniMax M2.7: DeepInfra at $0.25 input per 1M tokens (`deepinfra/MiniMaxAI/MiniMax-M2.7`)
- MiniMax M3: Weights & Biases at $0.23 input per 1M tokens (`wandb/MiniMaxAI/MiniMax-M3`)

## When to pick which

- MiniMax M2.7: Costs 17% less per input token: $0.25 vs $0.30 per 1M.
- MiniMax M2.7: Returns up to 204,800 tokens per response, vs 128,000.
- MiniMax M3: Larger context window: 1,000,000 vs 196,608 tokens.
- MiniMax M3: Supports image input, which MiniMax M2.7 doesn't list.

## FAQ

### Which is cheaper, MiniMax M2.7 or MiniMax M3?

MiniMax M2.7 is cheaper at 3 input tokens per output token: $0.4375 per 1M tokens blended ($0.25 input, $1.00 output), versus $0.525 ($0.30 and $1.20) for MiniMax M3, as of October 21, 2026.

### Which has the larger context window, MiniMax M2.7 or MiniMax M3?

MiniMax M3 has the larger context window: 1,000,000 tokens, versus 196,608 for MiniMax M2.7.

### Can I switch from MiniMax M2.7 to MiniMax M3 without changing my code?

Yes. Through LiteLLM both use the OpenAI format, so switching means changing the model name from "deepinfra/MiniMaxAI/MiniMax-M2.7" to "minimax/MiniMax-M3", or pointing a LiteLLM gateway model alias at the other model. You can also use one as a fallback for the other.