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

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.

MiniMax released M3.1 Flash weights on Oct 23 with 428B total and 23B active parameters, sparse attention in the first three full-attention layers and W4A4 MoE experts, claiming 100T tokens processed in 72 hours and confirming it was the anonymous Space Bunny.

M3.1 Flash weights are on Hugging Face and ModelScope: 428B total parameters with 23B activated per token, text/image/video input, 1M-token context, official 73.7% on DeepSWE v1.1 and 61.7 on OSWorld 2.0 under the MiniMax Community License.

After two weeks of free Space Bunny use in OpenCode for simple tasks, the author learned from MiniMax's blog that it was an early M3.1 Flash that processed over 200 trillion tokens anonymously; open weights are 428B total with 23B active per token.

M3.1 Flash weights are available in an October 23 repository, taking text, images and video and outputting text; the community license requires prominent attribution and prior written authorization above $20M yearly revenue.

MiniMax released M3.1 Flash weights at 428B total parameters with only 23B active per token, downloadable on Hugging Face and ModelScope; the poster sees sparse activation keeping inference costs down.

A license comparison notes MiniMax M3.1 Flash has 428B parameters with 23B active and open weights, but its community license requires a "Built with MiniMax" credit and written approval for companies earning over $20M a year.

The promised open weights arrived: 428B total parameters, 23B active per token, 1M context, text/image/video input, with day-0 vLLM support for Blackwell, Rubin and MI350.

On the day MiniMax released M3.1 Flash weights, g2c published a brief deriving the model's design and architecture from first principles.

A breakdown of the M3.1 Flash release: 428B total / 23B active MoE, text-image-video input, 1M context, ~252GB 8-bit download with SGLang/vLLM/TokenSpeed recipes, free for non-commercial use but requiring attribution and revenue-based notice or approval commercially.

A user speculates that M3.1 Flash shares M3's 428B total / 23B active parameters and is essentially M3 relabeled as a Flash tier, hinting a Pro model is coming.

A developer introduces MiniMax M3.1 Flash, formerly Space Bunny: a 428B-total/23B-active multimodal MoE with 1M context, QAT FP4 weights plus a DSpark draft head, runnable on SGLang and vLLM.