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

Unsloth's 2-bit dynamic GGUF shrinks the 428B M3.1 Flash to 143.9GB, runnable on a 150GB RAM machine with 91.5% top-1 agreement versus the original, plus official DeepSWE/OSWorld/CyberGym scores.

Using Unsloth's 2-bit GGUF (144GB) across 96GB VRAM plus system RAM, a custom forked engine hit 23–31 tok/s decode and 1120 tok/s prefill at 16K versus llama.cpp's 6.7 and 219 tok/s, with identical NLL and RAM bandwidth as the bottleneck.

Two DGX Sparks run the 428B M3.1 Flash at 2.15 bpw, hitting 33 tok/s on rewriting and 29 on code with 13.6s for an 8k prompt, but IFEval drops to 60.5% versus 89.0% for DeepSeek V4.1 Flash at 2.9 bpw, so a 3.06 bpw pack is being tested.

Unsloth's 2-bit dynamic GGUF shrinks the 428B MoE MiniMax M3.1 Flash to about 144GB with 91.5% top-1 token agreement in its own eval, runnable on 150GB of RAM.

Local inference project Project Maya now has the MiniMax-M3.1-Flash DSpark drafter read the whole conversation, making answers up to 1.5x faster with identical output, alongside a new web UI and AMD/MI50 support.

Unsloth's dynamic 2-bit GGUF shrinks M3.1 Flash to 143.9GB (45% smaller than official) while holding 91.5% Top-1 accuracy, runnable in 150GB RAM with setup steps included.

Local deployment: Maya-XS (93 GB) keeps every expert and Maya-S (122 GB) stays closest to the original, running on V100+ NVIDIA, Strix Halo and llama.cpp, plus a new loop guard against repeating answers.

Local deployment tool Project Maya v1.1.29 officially supports MiniMax-M3.1-Flash, and GGUF quantized builds are already available.

The post notes MiniMax M3.1 Flash is a 428B MoE that Unsloth's 2-bit dynamic GGUF shrinks to 143.9GB with 91.5% top-1 token agreement across seven domains, lowering the RAM barrier though not to laptop level.

A developer says a GGUF quant of M3.1 Flash is coming to Project-Maya and expects it to run faster than DeepSeek V4.1 Flash on an RTX 3090 given its smaller size.

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.

Unsloth published quantized GGUF builds for MiniMax-M3.1-Flash; the 2-bit version shrinks size by 45% while keeping 91.5% top-1 and runs in 150GB of RAM, with a deployment guide.