About MiniMax
The MiniMax token price index is $0.0370 per 1M tokens today, down 8.9% since yesterday, down 51.1% over the past week and down 63.1% over the past month. Over the past year it has ranged from $0.0370 to $0.209. It is the usage-weighted average of the posted prices of MiniMax's model at the mix of tokens agents send (87% cache read, 12% input, 1% output). It is calculated once a day, for each UTC day, shortly after the day ends.
MiniMax, a Shanghai AI company founded in December 2021 by former SenseTime researchers, makes the M-series language models, the Hailuo video models, speech models and the Talkie and Xingye character apps; Yan Junjie, a co-founder, is chief executive. It listed in Hong Kong on January 9, 2026 under stock code 0100. It publishes the weights of its M-series language models and some others, such as its H3 video model, on Hugging Face; those of M3 and M3.1 Flash are under the MiniMax Community License, which requires commercial users to credit MiniMax, and products with more than $20 million in yearly revenue to have its written authorization. Every MiniMax flagship from M2, released in October 2025, to M3 has been listed at $0.30 input and $1.20 output on its own API, M3 for prompts up to 512K tokens at half its $0.60/$2.40 list price, a launch discount that was permanent by June 9, 2026. On October 27, 2026, its index, $0.0417, was the third lowest of the seven Chinese companies on this site, above only DeepSeek's and Xiaomi's.
Models and prices
| Model | Input | Cache read | Output | At the mix |
|---|
| MiniMax-M3.1-Flash | $0.15 | $0.015 | $0.60 | $0.0370 |
| MiniMax-M3 | $0.30 | $0.06 | $1.20 | $0.100 |
USD per 1M tokens on October 28, 2026, from MiniMax's pricing overview; "At the mix" prices 87% of tokens as cache reads, 12% as input and 1% as output, as the index does. M3.1 Flash has one price for all prompt lengths. M3's row is for prompts up to 512K tokens, after a "Permanent 50% off" from $0.60/$2.40; longer prompts cost twice as much, and the Priority tier 1.5 times the standard rate. Cache reads cost 10% of the input price on M3.1 Flash and 20% on M3. MiniMax-M2.5 and M2.7 closed to new users on October 19, 2026 and retire on November 18.
What this index measures
Each day's value averages the prices MiniMax posts for its own models, weighted by each model's tokens in OpenRouter's daily rankings, with each model priced at the mix of tokens agents send: 87% cache reads, 12% input and 1% output, as OpenRouter measured in agentic use. Free variants are not counted. On October 27, 2026, M3.1 Flash carried 92.7% of MiniMax's counted tokens and M3 7.3%. MiniMax's own endpoint on OpenRouter, from which the index reads prices from October 28, 2026, listed the same prices that day.
The history starts on January 15, 2025, with MiniMax-01 at $0.20/$1.10; it had no cache price, so its cache reads are priced as input. On 195 days in 2025, the last on November 12, no MiniMax model the index counts was among OpenRouter's 50 most-used, so the value was carried: the last day's model shares, priced at that day's prices. June 8, 2026 is empty under the index's rules, because M3's price that day is not on record.
Other hosts serve the same weights: on October 28, 2026, OpenRouter listed 13 endpoints for M3.1 Flash, MiniMax's own among them, from $0.135 to $0.15 per million input tokens; 11, MiniMax's included, were at $0.15/$0.60. MiniMax published the weights on Hugging Face on October 23, four days after launch. The index uses MiniMax's own price.
What moves MiniMax's index
- Flash versus M3. At October 28, 2026 prices, M3.1 Flash costs 63% less than M3 at the mix; MiniMax puts its usage cost on agentic tasks "70% lower than its predecessor". From October 19 to 22, 2026, M3.1 Flash was among OpenRouter's 50 most-used only as its free variant, with up to 5.4 trillion tokens a day, and the index stayed at M3's $0.100. The paid model entered on October 23 with 39.0% of MiniMax's counted tokens and carried 92.7% on October 27, when the index was $0.0417, 58% below October 20. MiniMax changed no price.
- Cache prices. Every flagship from M2 to M3 has been listed at $0.30/$1.20 (M3 up to 512K tokens), but cache reads cost $0.03 on M2 to M2.5 and $0.06 from M2.7, as recorded in the index's price history, so M2.7 and M3 cost $0.100 at the mix against $0.0741. The index was $0.0741 from November 13, 2025 to March 17, 2026, between $0.0751 and $0.0960 from March 18 to May 31, 2026, as M2.7 and M2.5 shared usage, and $0.100 from June 12 to October 22, 2026.
- Free models. M2 was free on MiniMax's own API until November 12, 2025, after MiniMax extended an offer first set to end on November 7, and was not counted; the next day, at $0.30/$1.20, the index fell to $0.0741 from $0.209. Free variants on OpenRouter carried more tokens than MiniMax's paid models on 34 days, most recently October 19 to 22, 2026: on September 4, 2026, M3's free variant carried 929 billion tokens, more than four times paid M3's 211 billion.
Price history
- January 15, 2025: MiniMax-01 released with open weights at $0.20/$1.10.
- October 27, 2025: MiniMax-M2 released, free on MiniMax's API until November 12 (the release notes had given November 7); then $0.30/$1.20 with cache reads at $0.03, on MiniMax's November 2025 price list.
- February 12, 2026: MiniMax-M2.5, at $0.30/$1.20 with cache reads at $0.03 on MiniMax's February 13 price list, as M2.1 on its January 2, 2026 list.
- March 18, 2026: MiniMax-M2.7; MiniMax's April 14 price list, the first with its dollar prices, gives $0.30/$1.20 with cache reads at $0.06.
- June 1, 2026: MiniMax-M3 released, listed at $0.60/$2.40 with a "7-day 50% off" to $0.30/$1.20 that by June 9 was "Permanent 50% off".
- October 19, 2026: MiniMax M3.1 Flash at $0.15/$0.60, with cache reads at $0.015.
FAQ
How much does the MiniMax API cost per million tokens? On October 28, 2026, MiniMax M3.1 Flash cost $0.15 input and $0.60 output, and M3 $0.30 and $1.20 for prompts up to 512K tokens, as in the table above.
Why did MiniMax's index fall 58% in a week in October 2026? From October 20 to 27, 2026, usage moved to M3.1 Flash, which costs 63% less than M3 at the mix and carried 92.7% of MiniMax's counted tokens on October 27. MiniMax did not change its prices.
How often is it updated? Once a day: each UTC day's value is calculated in the first update after that day ends, from the day's OpenRouter volumes and the prices posted then. Updates are scheduled every 2 hours but can start hours late.
Sources
- MiniMax: pricing overview
- MiniMax: model list and M2.5 and M2.7 retirement notice
- MiniMax: model release notes
- MiniMax: M3.1 Flash announcement, October 19, 2026
- Hugging Face: MiniMaxAI models
- TechNode: MiniMax releases M3.1 Flash's weights, October 23, 2026
- OpenRouter: Inferred 03, Cache is Crucial, October 8, 2026, the token mix
- OpenRouter: rankings, usage weights, CC BY 4.0
Text reviewed Oct 28, 2026; the figures in it are as of that date. The figures at the top of this page update once a day.
Cite as: Compute Screener, “MiniMax token price index”, computescreener.com/tokens/minimax. The figures are free to reuse under CC BY 4.0 with credit to Compute Screener.
