MiniMax-M3.1-Flash vs Qwen3.5-122B-A10B

MiniMax-M3.1-Flash

A 428B mixture-of-experts model, 23B active, for agent work, coding and computer use, with image and video input, about 1M tokens of context and parallel tool calls. It needs a 4 to 8 GPU server for 250 GB of NVFP4 and MXFP8 weights, and vLLM and SGLang support was pre-release on 24 October 2026. Commercial use needs attribution and notice.

Qwen3.5-122B-A10B

A strong mid-size cousin for coding, agent steps and reading images or video, with thinking you can switch off per request. Alibaba's 4-bit build needs about 95 GB, which fits one 128 GB unified-memory machine, and several providers host it. The main risk is silence, with no safety results, agent guidance or answers from Alibaba.

MiniMax-M3.1-Flash

1. Check the use case first; commercial products must show "Built with MiniMax M3.1-Flash" and notify MiniMax once, or get written authorisation above $20M yearly revenue
2. On 8 H100 or H200 GPUs use vLLM's nightly image vllm/vllm-openai:minimax-m31-flash with --tensor-parallel-size 8 --block-size 128 --kv-cache-dtype fp8 --moe-backend marlin --max-model-len 262144
3. Pass --reasoning-parser minimax_m3 --tool-call-parser minimax_m3 --enable-auto-tool-choice in vLLM, or minimax-m3 for both in SGLang with --trust-remote-code
4. --block-size 128 is mandatory in vLLM, since MSA's sparse block size is 128
5. Set effort per request with chat_template_kwargs reasoning_effort (low, medium, high, xhigh, max); sample at temperature 1.0 and top_p 0.95
6. On Blackwell or MI355X, 4 GPUs serve it with the DSpark draft from dspark/; vLLM's Hopper command leaves DSpark off and turns video input off
7. For GGUF builds by others, use llama.cpp b11861 or later; earlier builds mis-scale the NVFP4 experts and produce garbage
8. Without the hardware, call MiniMax-M3.1-Flash on MiniMax's API (minimax-api), whose thinking defaults to max effort

Qwen3.5-122B-A10B

1. Quickest start on a 128 GB machine is ollama run qwen3.5:122b (Q4_K_M, 81 GB of files, about 97 GB with context)
2. Alibaba's 4-bit build serves with vllm serve Qwen/Qwen3.5-122B-A10B-GPTQ-Int4 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder --quantization moe_wna16, though that kernel isn't documented for AMD gfx1151
3. Use the qwen3_coder tool parser and the qwen3 reasoning parser, as the card's commands do
4. Turn thinking off with "chat_template_kwargs": {"enable_thinking": false} in the request body, since it's on by default and /think and /nothink aren't supported
5. Sampling per the card is temperature 0.6, top_p 0.95 and top_k 20 when thinking, and temperature 0.7, top_p 0.8, top_k 20 and presence_penalty 1.5 without
6. Keep the context at 128K or more if you rely on thinking, and turn on YaRN only past 262,144 tokens
7. Add --language-model-only in vLLM to skip the vision encoder and free memory for the KV cache when you only need text
8. Validate any JSON it returns, since structured output isn't documented