Moving from M2 to the MiniMax M2.1 API tightened our agent loops without a rewrite.
'%3e%3c/path%3e%3cdefs%3e%3clinearGradient%20gradientUnits='userSpaceOnUse'%20id='lobe-icons-hailuo-_R_0_'%20x1='.539'%20x2='27.487'%20y1='.884'%20y2='27.022'%3e%3cstop%20offset='.09'%20stop-color='%23FFAB0C'%3e%3c/stop%3e%3cstop%20offset='.31'%20stop-color='%23FF5538'%3e%3c/stop%3e%3cstop%20offset='.46'%20stop-color='%23E9405D'%3e%3c/stop%3e%3cstop%20offset='.75'%20stop-color='%23D266DA'%3e%3c/stop%3e%3cstop%20offset='.89'%20stop-color='%23D584EF'%3e%3c/stop%3e%3c/linearGradient%3e%3c/defs%3e%3c/svg%3e)
MiniMax M2.1 is the point-release refinement of M2 — the same agent-and-coding focus, tightened, with sharper tool-loop reliability and coding at the same low price. Served on RouterBase over an OpenAI-compatible endpoint at $0.30 / 1M input and $1.20 / 1M output, with cached input at just $0.03, 5% under list; if M2 is already in your stack, it is a drop-in step up.
Input
MiniMax M2.1
Hi! I'm a helpful AI assistant. What can I do for you?
MiniMax M2.1 API — The Refined M2 for Agents and Coding
The MiniMax M2.1 API serves MiniMax M2.1, the point-release refinement of M2 — sharper agentic tool loops and coding at the same low price.
MiniMax M2.1 is the refined follow-up to M2: the same agent-and-coding focus, tightened. The MiniMax M2.1 API keeps the fast, cheap profile that made M2 a practical default — $0.30 per 1M input tokens and $1.20 per 1M output, with cached input at just $0.03 — while sharpening tool-loop reliability and coding. If M2 is already in your stack, the MiniMax M2.1 API is a drop-in step up; if you are starting fresh, it is the newer of the two to build on.
You reach the MiniMax M2.1 API through RouterBase on the OpenAI chat-completions protocol; set the model to minimax-m2-1 and your client is unchanged. The MiniMax M2.1 API bills $0.30 per 1M input, $1.20 per 1M output, and $0.03 for cached input — 5% under list, on one key that reaches the whole catalog.
What the MiniMax M2.1 API gives you
Same agent-and-coding profile, tightened.
Refined M2
The MiniMax M2.1 API is the point-release refinement of M2 — same focus, tightened.
Sharper agent loops
Tighter tool-loop behavior makes the MiniMax M2.1 API a cleaner driver for multi-step agents.
Strong at coding
Writing and fixing code is still front and center for the MiniMax M2.1 API.
Same low pricing
At $0.30 in and $1.20 out, it costs the same as M2 — 5% below list.
Near-free cached reads
Cached input is just $0.03 per 1M, so repeated context on it costs almost nothing.
Drop-in from M2
Already on M2? Point at the MiniMax M2.1 API by changing one string — the OpenAI shape is unchanged.
Get going
First call to the MiniMax M2.1 API in three steps
Quick to wire, ready to drive a tool loop.
Create a key
One RouterBase key reaches the MiniMax M2.1 API and every other model in the catalog.
Point and name
Send any OpenAI client to routerbase.com/v1 with model=minimax/minimax-m2.1. Streaming and function calling work out of the box.
Wire your tools
Give the MiniMax M2.1 API your function schemas and let it drive the loop; usage comes back per response so cost stays visible.
Pay only for what you use
RouterBase passes through partner-tier pricing. Compared against the model's official published API rate.
Teams building on the MiniMax M2.1 API
The refined agent-and-coding model, running cheap.
M2.1 drove our multi-step tasks more cleanly than the version before it.
Same $0.30 in, sharper agents — the MiniMax M2.1 API was an easy upgrade.
Cached reads at $0.03 kept our long loops on the MiniMax M2.1 API basically free.
Function calling held up, so our whole tool layer runs on the MiniMax M2.1 API.
Our coding agent got noticeably steadier on M2.1.
One string changed from M2; the MiniMax M2.1 API kept everything working and cleaned up the tool calls.
We default agent and coding traffic to the MiniMax M2.1 API and keep it all on one key.
The MiniMax M2.1 API was fast enough to sit inside our live coding assistant, same as M2.
MiniMax M2.1 API — common questions
What to know before you route traffic to it.
RouterBase's hosted access to MiniMax M2.1, the refined point-release of M2 for agentic workflows and coding, over an OpenAI-compatible endpoint.