See how ChatGPT, Perplexity and Google AI Overviews describe you today
Free AI Visibility AuditMiniMax M2 is a text-only large language model from MiniMax, built for end-to-end coding and agentic workflows and priced low enough to run agent loops at volume.
01 / decision
Each figure sits next to the middle of the field, so it can be read as dear or cheap, wide or narrow, rather than floating on its own. Compared against the 126 models we track, not against an absolute standard.
Capability
Not measured
No published benchmark scores yet
Input / 1M tokens
$0.30
field median $0.30
Output / 1M tokens
$1.20
field median $1.25
Context
205K
field median 524K
02 / overview
Four questions, answered separately, because somebody arrives at one of them rather than at the top of the page.
The job it was built for, and the job it is not for.
MiniMax M2 is a sparse large language model that activates a small fraction of its total parameters on each pass, which is what keeps it cheap and quick to serve. It was built for coding work that runs start to finish and for agents that chain many steps together, rather than for open-ended chat. It handles text only, so image, audio and video work needs a different model.
Where it sits in its line, and when a sibling is the better pick.
MiniMax M2 first appeared in October 2025 as MiniMax's compact, efficiency-led option rather than a maximum-capability flagship. It is the right pick when a task runs the same prompt shape thousands of times, or when an agent takes many turns to finish a job and the token bill compounds. It is the wrong pick when you need multimodal input, or when a single hard reasoning problem justifies paying for a frontier model.
The API, the apps it powers, and what its limits let you do.
MiniMax M2 is reached through the MiniMax API and through the usual third-party inference and routing platforms that carry open-weight and low-cost models. Its context window holds a large codebase, a long tool-call transcript or a stack of retrieved documents in one request, and its output ceiling is unusually generous, so it can write long files or full multi-step plans without being cut off mid-answer. Everything in and out is text, which suits code, tool calls and structured output.
What changes because this exists, or why it does not.
Agentic systems fail on economics more often than on capability, because every retry, every tool result and every re-read of context is billed again. MiniMax M2 makes the long-loop pattern affordable to leave running, so teams can build agents that take twenty steps instead of rationing them to five. Nothing here is novel in kind, it is the cost curve moving under work that was already possible.
03 / evidence
Split by category, so a strong number never hides a thin evidence base. Verified means we read it on the benchmark's own published results; a provider's claim about its own model is shown and labelled rather than dropped.
No published benchmark scores for this model yet.
We publish a score only where we can link the result to where it was published. Until a benchmark result for this model exists in a source we read, this section stays empty rather than being filled with a provider's marketing figure.
How we decide what counts as evidence04 / ledger
Every published row, grouped by category, each compared with the best published score on the same benchmark. 'Is 64% good' is a question nobody can answer; '26 points behind the leader' is one anybody can.
Nothing in the ledger yet.
Each row here carries a score, the benchmark it came from, what the leading model scored on the same test, and a link to the published result. Rows appear as results are published and read.
How we decide what counts as evidence05 / capability
Where this model is strong, and against how many peers. Ranks are against models with evidence in that category, not against everything we track: ranking against models nobody tested would rank who published, not who is better.
No category scores to shape yet.
A category score is the weighted mean of the benchmarks published for it. With no published rows there is nothing to average, and an empty chart drawn at zero would say something false.
How we decide what counts as evidence06 / cost
List API rates as last read from the provider, with the source on every row, plus every change we have recorded since we started tracking it.
| Charge | Price | Unit | Read on | Source |
|---|---|---|---|---|
| Input | $0.30 | per 1M tokens | 2026-09-24 | Check |
| Output | $1.20 | per 1M tokens | 2026-09-24 | Check |
MiniMax M2 sits at the low end of the market, closer to small open-weight models than to frontier systems, and the gap is wide enough to change what is worth building. Output costs several times what input does, so the savings are largest on retrieval-heavy and tool-heavy work where most tokens go in rather than out.
| Workload | Input / month | Output / month | Cost |
|---|---|---|---|
| A small product team | 20M tokens | 5M tokens | $12.00 |
| A busy support assistant | 200M tokens | 40M tokens | $108.00 |
| A document pipeline | 1000M tokens | 100M tokens | $420.00 |
List API rates, no caching and no batch discount, which both providers offer and which change the answer a great deal. Treat these as the ceiling, not the bill.
07 / specs
As listed by the provider's own catalogue and re-read every few hours. Anything absent is absent there too.
| Context window | 204,800 tokens |
|---|---|
| Maximum output | 176,947 tokens |
| Modalities | text |
| Released | 23/10/2025 |
| Status | Current |
| Catalogue identifier | minimax/minimax-m2 |
08 / lineage
What this model replaced, what replaced it, and what else its provider has in the field.
Also from MiniMax
09 / line
Every model in this family in release order, so a page from eight months ago says in one glance that two newer ones exist.
Nothing else in this line yet.
A line is worked out from the naming and the release dates across every model page we hold. It fills in as the provider ships successors, or as we pick up the models that came before this one.
How we decide what counts as evidence10 / notes
What this model changes for a brand trying to be cited in AI answers, and every change we have logged since it launched.
What it changes for you
A cheap agentic model means more retrieval passes, not fewer. When an agent can afford to read ten sources instead of two before answering, thin content gets skipped and the pages with specific, checkable detail get pulled in. For a SaaS brand, that rewards documentation, pricing pages and comparison content that survive being read closely rather than skimmed.
Most buyers meet MiniMax M2 indirectly rather than by choosing it. It sits behind coding assistants, agent frameworks and internal tools where the builder picked it on price, and it is available through the MiniMax API and the model routers that developers default to. There is no major consumer AI search surface running on it, so it shapes what gets built more than what gets read.
Nothing published here yet. Changes appear within hours of a provider announcing them.
11 / questions
The things people ask about this model, answered from what is on this page rather than from anywhere else.
Models change what gets cited. We measure whether AI answers name your brand or your competitors across every assistant, and show you what to change.