MiniCPM-o 4.5 vs Qwen3.8-Flash-Next

MiniCPM-o 4.5

A small omni cousin that sees, listens and talks back, with full-duplex speech and voice cloning, in about 7 GB at 4-bit through llama.cpp or Ollama for text and images. Speech output needs OpenBMB's remote code or llama.cpp-omni, and 28 GB of GPU for the full demo. The main risks are repository code and pickle files and no safety results.

Qwen3.8-Flash-Next

A large sparse cousin with 6B active parameters for agentic coding and multimodal work, previewing Qwen's next architecture. It needs a 128 GB Mac for Ollama's MLX build or about 144 GB for the 4-bit GGUF, and Qwen Cloud serves it. The main risk is a licence that needs a separate deal for hosted-model and coding or office assistant businesses.

MiniCPM-o 4.5

1. Quickest start for text and images is ollama run openbmb/minicpm-o4.5, about 6 GB
2. With vLLM 0.16.0 or later pass --trust-remote-code, and send stop_token_ids [151643, 151645] as OpenBMB's guide does
3. vLLM, llama.cpp and Ollama return text only; for speech out use the card's Transformers 4.51.0 code or OpenBMB's llama.cpp-omni
4. In llama.cpp pass both the language GGUF and MiniCPM-o-4_5-vision-F16.gguf with --mmproj, and use --jinja --reasoning-budget 0 to turn thinking off
5. Keep prompts within 40,960 tokens
6. Don't rely on tool calls or JSON output, neither is documented; validate any you get

Qwen3.8-Flash-Next

1. On a 128 GB Mac start it with ollama run qwen3.8-flash-next:125b-mlx (105 GB, NVFP4), on Linux the smallest Ollama build is qwen3.8-flash-next:125b-a6b-q4_K_M at 120 GB, which won't fit a 107.4 GB GPU share
2. On a server, vllm serve Qwen/Qwen3.8-Flash-Next-FP8 --tensor-parallel-size 4 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder
3. Use the qwen3_coder tool parser and the qwen3 reasoning parser, as Qwen's GitHub README does
4. Thinking is on by default, pass chat_template_kwargs with enable_thinking false to skip it, and read the reasoning from reasoning_content, even though SGLang's cookbook says thinking can't be turned off
5. reasoning_effort takes xhigh (default), medium or low, and any other value makes the template raise an error
6. Send at most one system message and put it first, since the template raises an error otherwise
7. Sampling per the card is temperature 1.0 and top_p 0.95 when thinking, temperature 0.7, top_p 0.8 and presence_penalty 1.5 without
8. vLLM keeps the 51B n-gram table in host memory on 80 GB GPUs by default, per its recipe
9. Validate JSON output yourself or use the runtime's guided decoding, since structured output isn't documented
10. Read the Qwen Community License before serving it to others as an API or inside a coding or office assistant product