MiniMax review

MiniMax is an AI company that develops foundation models and sells AI products and API access. The company says its proprietary models cover text and multimodal tasks, but the published pricing details do not publish text API rates or usage limits.

Its product lineup includes MiniMax Code, described as an all-in-one agent built for MiniMax models, and MiniMax Design. The model list includes M3.1 Flash, positioned by MiniMax for agent performance and speed; H3, described as an open-weight multimodal model; and Speech 2.8 and Music 3.0. Developers and enterprises can also use MiniMax’s API platform.

The published plans are for audio and video, not a general text-model subscription. Audio starts at $5 per month, or $4 per month with annual billing. Video packages start at $1,000 and are valid for one month. The pricing information checked on November 6, 2026 lists no free plan; a free trial is not specified.

The clearest advantage is range: MiniMax offers text and multimodal models alongside speech, music, video, coding and design products. That gives teams more than one way to try the company’s models, though it does not tell us how well each product fits a particular workflow.

For speech API buyers, the plans spell out concrete allowances. Audio Starter includes 100,000 points, 10 voice slots and 10 requests per minute (RPM) for $5 monthly, or $4 per month with annual billing. Higher tiers raise those limits: Audio Pro lists 1.1 million points, 250 voice slots and 200 RPM for $99 monthly, or $79.17 per month annually. All listed audio plans support all speech models. Three-month billing is listed as 10% off, while annual billing is 20% off.

MiniMax positions M3.1 Flash around agent performance and speed, and describes H3 as open-weight. In one Reddit thread, a user said H3 held up well against a closed video model; commenters disagreed about the quality, including noting changes in the subject’s appearance. That is a useful sign of interest, not a controlled comparison.

The biggest buying gap is text API pricing. The published pricing page details audio and video packages, but gives no text-model price, token allowance or rate limit. We would not choose a text model for production on the strength of a model lineup alone; confirm the relevant API rate limits and costs first.

Video is expensive by the published package prices. Video Standard costs $1,000 for 3,760 points at 20 RPM; Pro is $2,500 for 9,920 points at 30 RPM; and Scale is $4,500 for 18,900 points at 40 RPM. Each package is valid for one month. Custom pricing is available, but the price is not published.

Audio has more accessible entry pricing, but it still comes with caps: Starter is limited to 10 voice slots and 10 RPM, while even Business lists 800 voice slots and 800 RPM. Custom audio pricing offers points as needed and unlimited RPM/TPM, but MiniMax does not publish its cost. User reports are mixed too: one Reddit thread praised an M2 result on technical problems, while commenters reported tool-call formatting issues and weak performance outside STEM. These are individual reports, not a reliable measure of current models.

MiniMax’s listed entry price is $4/month, but that is Audio Starter’s annual-billing rate; month-to-month costs $5. Starter includes 100,000 audio points, 10 voice slots and 10 RPM. Audio tiers rise quickly: Standard is $30 monthly ($24 annual) for 300,000 points; Pro is $99 ($79.17 annual) for 1.1 million; Scale is $249 ($199.17 annual) for 3.3 million; and Business is $999 ($799.17 annual) for 20 million. RPM and voice-slot caps also increase by tier, and all five plans support the speech models.

Video is a much larger commitment: Standard costs $1,000 for 3,760 points, Pro $2,500 for 9,920, Scale $4,500 for 18,900, and Business $6,000 for 26,780. Each package is valid for one month. Audio offers monthly, three-month billing at 10% off, or annual billing at 20% off; the listed annual prices reflect the discount. Custom audio and video pricing is available, but add-ons are not listed. Text API pricing is not published on the vendor's site.

The published material does not document import tools, data-export formats, compatibility layers or migration support, so we can’t give a reliable estimate of the work involved in switching. Before moving workloads, inventory model-specific prompts, speech workflows and any integrations that depend on MiniMax’s API. MiniMax identifies H3 as open-weight, but the published details do not describe export options for other models or explain how to transfer stored assets.

There is no Reddit post or comment evidence in the published material, so we can’t attribute reasons for leaving to MiniMax users. The pricing details do show two potential reasons to reassess: text API pricing is not published, and video packages start at $1,000 and expire after one month. If you are leaving, confirm that your replacement supports the model behaviors you rely on and check MiniMax’s current data-export terms before canceling; those terms are not included on the vendor's site.