MiniMax M3 · MiniMax

What does MiniMax M3 cost to run in a coding agent?

MiniMax M3 costs about $0.013 per average coding-agent request on our measured traffic, at OpenRouter's listed prices of $0.30 per million input and $1.20 per million output tokens. It scores 29 on the Artificial Analysis Intelligence Index, 31st of the 48 models we track. Source: Artificial Analysis (artificialanalysis.ai)

List prices: OpenRouter listed price, checked 2026-10-13. Request costs: Standard Compute gateway data, 79,301 coding-agent requests from 75 users, Sep 13 to Oct 12, 2026. How we compute this

Run MiniMax M3 in your agent

Or let the router send each step to the cheapest model that can do it, on providers we vet for price, uptime and speed. Plans from $19 a month.

See plans

What MiniMax M3 costs per coding-agent request

Cost at list price	MiniMax M3
Prompt under 100K tokens50% of requests, 87% of input cached	$0.0060
Prompt of 100K to 272K39% of requests, 88% of input cached	$0.016
Prompt over 272K tokens11% of requests, 86% of input cached	$0.035
Average request	$0.013
100 requests	$1.29
Median session16 requests; a session ends after 10 idle minutes	$0.16
A month of regular usemedian of 20 users active on 5+ days	$20.98

Each logged request priced at the listed rates above for its own token counts, including cache reads and any higher rate for long prompts. Average output: 698 tokens per request.

MiniMax M3 benchmarks, speed and latency

Artificial Analysis	Value	Rank
Intelligence Index10 evaluations, 0 to 100	29	31 of 48
Terminal-Bench 4.0agentic coding in a terminal	2.0%	27 of 39
Output speedmedian	119 tokens/s	14 of 44
Time to first answer tokenincludes thinking time	18 s	8 of 44

Source: Artificial Analysis (artificialanalysis.ai). Measured variant: MiniMax-M3, speed on MiniMax. Fetched October 13, 2026. Rank is among the models on this site, older versions included. Artificial Analysis runs Terminal-Bench 4.0 in its own setup, so its scores differ from the tbench.ai leaderboard.

About MiniMax M3

MiniMax M3 is the June 2026 model in MiniMax's agent-focused open-weight line, with a 1M-token context. MiniMax pitches the line for tool use and long-horizon tasks, but independent results are modest: 29 on the Artificial Analysis Intelligence Index and 2.0% in its Terminal-Bench 4.0 run, at about 118 output tokens per second. At OpenRouter's listed price it costs $0.013 per request in our measurements, about twice DeepSeek V4.1 Flash at peak rates, which scores higher on both boards.

Sources: Artificial Analysis

MiniMax M3 FAQ

How much does MiniMax M3 cost per coding-agent request?

$0.013 on average at OpenRouter's listed price, $1.29 per 100 requests, and $0.16 for a median 16-request session. That is 79,301 real coding-agent requests (Sep 13 to Oct 12, 2026) priced at OpenRouter's listed rates; the median request sent 99,534 input tokens, 87% of input was read from cache, and requests returned 698 output tokens on average.

Does MiniMax M3 charge more for long prompts?

No. MiniMax M3 bills the same rate per token up to its 1M-token context window.

How fast is MiniMax M3?

Artificial Analysis measures MiniMax-M3 at 119 output tokens per second, with 18 seconds to the first answer token, thinking time included, on MiniMax's API. Source: Artificial Analysis (artificialanalysis.ai).

Is MiniMax M3 good for AI agents?

Agents that need an open-weight model with a 1M context. On our scoring, which is calibrated to public benchmarks, it rates 5.0/10 for agentic ability and 6.3/10 for output quality. Open weights.

What is MiniMax M3's context window?

MiniMax M3 has a 1.0M-token context window, large enough to hold sizeable codebases or document sets in a single run. MiniMax is the maker.

Is MiniMax M3 open source?

Yes. MiniMax M3 ships with open weights, so you can self-host it or run it through a hosted provider such as OpenRouter. Check the license before commercial use.

What are the downsides of MiniMax M3?

The main trade-offs: Weak on long agentic tasks (2.0% in Artificial Analysis's Terminal-Bench 4.0 run); DeepSeek V4.1 Flash scores higher for about half the cost per request. It's strongest for agents that need an open-weight model with a 1M context.