MiniMax M3.1 Flash in practice: tests and comparisons with DeepSeek V4.1 Flash

The author recommends M3.1 Flash for boring background agent jobs but not near production code without checking, citing $0.15 input, $0.60 output and $0.015 cached input per 1M tokens.

MiniMax confirms Space Bunny was an early M3.1 Flash, priced at $0.15 input and $0.60 output per million tokens, and that it processed about 56.7 trillion tokens anonymously on OpenRouter.

A developer demos FlashDesk, a customer-support agent built with MiniMax M3.1 Flash that triages tickets and drafts replies, but cites only MiniMax's own speed and cost claims without independent results.

The author breaks down Unsloth's quantization table (1-bit 129GB, 2-bit ~150GB, 4-bit 249GB), estimates a ~87 tok/s theoretical ceiling for 2-bit at 91.5% top-1, and warns the 1M context can't be extrapolated from the 16K table.

A price table of the 25 most-used OpenRouter models by tokens Oct 19-25 puts MiniMax M3.1 Flash at $0.60 per 1M output tokens, level with DeepSeek V4.1 Flash 1017, across a 150x spread from $0.13 to $20.

On one video prompt, MiniMax M3.1 Flash cost $0.21 and produced bouncing debris, rising embers and trees burning one by one, while Kimi K3.1 Pro cost $6.10 for blocky cubes and solid lava — 29x the price for a worse result.

Given a small Python shop library with five planted bugs and one slow function, M3.1 Flash fixed all six with no hints, found a seventh, and cut a duplicate check from 23.8s to 0.01s, though it also changed unrequested behavior.

A user reports the same task consumed 30% of a 5-hour quota on GPT-6.1 Sol but exhausted three full 5-hour quotas on MiniMax M3.1 Flash.

On the same 3D Earth scene prompt, M3.1 Flash finished in 15 minutes for free during the promo while GPT-6.1 Sol took an hour and $3.20, with the author preferring M3.1 Flash's visuals and interactivity.

On a TP=2 Spark setup M3.1 Flash gets 29.4 tok/s on code and 27.8 on tool calls at 131K context, versus roughly 90 for DeepSeek, because it reads about 7.7 GB per token against DeepSeek's 3.8 GB.

The author compares MiniMax M3.1 Flash at ¥1 per million input and ¥4 output tokens with DeepSeek's Flash update two days earlier, arguing the race has shifted from smartest to cheapest model that finishes the job.

On the same one-shot kaiju prompt, Opus 5.5 max cost $136.70 over 11h27m, Gemini 4 Argon high about $4.30 in 48 minutes, DeepSeek V4.1 Flash 1017 max $0.94 over 3h11m, and MiniMax M3.1 Flash max about $2.28 over 8h44m.