Kimi K3 Low vs MiniMax-M3: cost and benchmark comparison

On the headline score, Kimi's Kimi K3 (Low) and MiniMax's MiniMax-M3 are almost level. The figures come from tracked catalogue data, not from our own testing, so treat them as a starting point for a trial on your tasks. Cost is where they split the most, with MiniMax-M3 working out at about 9% of Kimi K3 Low's blended rate.

Token pricing and a sample bill

The list prices are $3.00 in and $15.00 out for Kimi K3 Low, and $0.30 in and $1.20 out for MiniMax-M3, per million tokens. For output tokens, Kimi K3 Low is 13 times the price of MiniMax-M3.

A reference job of 1 million input tokens and 200,000 output tokens costs $6.00 with Kimi K3 Low and $0.54 with MiniMax-M3, so MiniMax-M3 saves $5.46. Scale it up to a chat app with 50,000 sessions of 800 input and 300 output tokens and the bill is $345 for Kimi K3 Low against $30.00 for MiniMax-M3.

Benchmark scores

The catalogue puts Kimi K3 Low at 17.4 and MiniMax-M3 at 18.9 on its intelligence index. It is a small margin for an index that averages many evaluations. In relative terms, MiniMax-M3 scores 9% higher.

For context, that places Kimi K3 Low 86th and MiniMax-M3 79th out of 109 scored models still on offer.

There are no coding or math index scores for either model in the catalogue, which keeps this comparison at the aggregate level.

Cost per point of intelligence

A blunt value measure, blended price over intelligence score, puts Kimi K3 Low at $0.34 and MiniMax-M3 at $0.028 per point. With the higher score and the lower cost per point, MiniMax-M3 is the better value per point on these numbers.

Speed and latency

Measured output speed is 49 tokens per second for Kimi K3 Low and 121 tokens per second for MiniMax-M3. MiniMax-M3 is ahead by 144% on throughput. The wait for the first token averages 3.6 seconds on Kimi K3 Low against 0.82 seconds on MiniMax-M3. Kimi K3 Low waits 4.4 times as long before its first token.

Add the two numbers up for a 400-token reply and MiniMax-M3 finishes first: 12 seconds for Kimi K3 Low, 4.1 seconds for MiniMax-M3.

Which is newer

Kimi K3 Low arrived on July 16, 2026, about a month after MiniMax-M3 (June 1, 2026). That is close enough to count them as the same generation.