MiniMax M3 vs Kimi K3

MiniMax M3 and Kimi K3 both read text, images and video and are presented for long agent work and coding. Kimi K3, at 2.8 trillion parameters, is a very large model, and both have published weights.

Today's prices and specs

Prices are in Rial per million tokens and specs are read straight from the IranRouter catalogue. The best value in each row is green. The intelligence and coding indexes come from public benchmarks, and hovering over a speed or first-token figure shows where it was measured.

MiniMax: MiniMax M3MinimaxMoonshotAI: Kimi K3Moonshot AI
Input (IRR / 1M tokens)823,806527,236
Output (IRR / 1M tokens)3,295,22543,936,328
One message (IRR)3,29523,023
Context (tokens)1M1M
Max output (tokens)512K944K
Intelligence18.917.4
Coding——
Speed (tokens/s)71From public market statistics for this model72From public market statistics for this model
First token (s)1.0From public market statistics for this model1.2From public market statistics for this model
Health, 30 days87.88%93.73%
Usage rank#44#38
Can doImage inputTool callingJSON outputReasoningImage inputTool callingJSON outputReasoning
ReleasedMay 31, 2026Jul 16, 2026

The differences at a glance

  • MiniMax M3's input price is about 1.6 times Kimi K3's.
  • Kimi K3's output price is about 13.3 times MiniMax M3's.
  • One typical message costs about 3,295 IRR on MiniMax M3 and about 23K IRR on Kimi K3.
  • Both hold about 1M tokens of context.
  • Both accept text, images and video as input.

When to choose MiniMax M3

  • Multimodal agent work that has to stay on one goal for a long time.
  • Coding with image or video input.

When to choose Kimi K3

  • Complex coding and multi-step knowledge work.
  • Long-running agents where reasoning is the main part of the job.

Before you decide

A larger model does not by itself guarantee a better result. Run one complete task on each and look at both the output and its cost in the dashboard.

Python
from openai import OpenAI

client = OpenAI(api_key="ir-...", base_url="https://iranrouter.com/v1")

for model in ["minimax/minimax-m3", "moonshotai/kimi-k3"]:
    r = client.chat.completions.create(model=model, messages=[{"role": "user", "content": "..."}])
    print(model, r.choices[0].message.content)

Frequently asked questions

Which is cheaper, MiniMax M3 or Kimi K3?

It depends on the work: at today's rates Kimi K3 has the cheaper input and MiniMax M3 the cheaper output. Long prompts with short answers lean on input, long answers on output.

Which suits video?

In today's catalogue both accept video as input. Few people have long experience of this yet, so start with a short sample.

Can I try both on one account?

Yes. One ir- key and the same Rial wallet reach both, and switching means changing only the model value in the request. You can set a spending limit on each key.

Get connected today

Create an account, top up your wallet and make your first request in minutes — no contract, no foreign card.

Create a free accounthttps://iranrouter.com/v1