Nemotron 3 Super 120B A12B vs Qwen3-VL-30B-A3B-Instruct

Nemotron 3 Super 120B A12B

A long-context agent cousin with trained parallel tool calls, a reasoning switch and a full account of its training data, under a free commercial licence. It needs about 100 GB of GPU memory, a 128 GB unified machine or Blackwell hardware for NVIDIA's NVFP4. The main risks are serving commands that run repository code and no published safety results.

Qwen3-VL-30B-A3B-Instruct

A mid-sized vision-language cousin for screenshots, documents, charts and video that runs fast for its size, with about 3B active parameters. It needs about 23 GB for Alibaba's Q4_K_M GGUF with its vision file, loads in every major runtime under plain Apache 2.0, and three providers host it. With no published safety work, add guardrails before it drives a GUI.

Nemotron 3 Super 120B A12B

1. On the reference machine start it with ollama run nemotron-3-super, the 87 GB q4_K_M build, the only one that fits (about 104 GB with the margin)
2. Raise the Strix Halo GPU memory limit first with amd-ttm --set 100, since the default of about 62.8 GiB is far below what it needs
3. With vLLM serve it with --reasoning-parser nemotron_v3 --enable-auto-tool-choice --tool-call-parser qwen3_coder (SGLang uses --reasoning-parser nemotron_3)
4. Use temperature=1.0 and top_p=0.95 for every task, as the card says
5. Thinking is on by default. Pass enable_thinking: false or low_effort: true in chat_template_kwargs for short tasks
6. For coding agents pass force_nonempty_content: true in chat_template_kwargs, so a cut-off reasoning trace still returns content
7. Raise context past 262,144 only with the card's flags (VLLM_ALLOW_LONG_MAX_MODEL_LEN=1, --max-model-len 1048576)

Qwen3-VL-30B-A3B-Instruct

1. Start it with llama-server -hf Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF:Q4_K_M --jinja -c 32768, or ollama run qwen3-vl:30b-a3b-instruct
2. In Ollama the plain qwen3-vl:30b tag is the Thinking edition, so name 30b-a3b-instruct
3. With the GGUF, load the mmproj vision file too, or images won't be read
4. Parse tool calls as JSON inside <tool_call> tags, or use SGLang's --tool-call-parser qwen
5. Use the shipped sampling, temperature 0.7, top_p 0.8 and top_k 20, from generation_config.json
6. Validate JSON output yourself, since structured output isn't documented
7. vLLM needs 0.11.0 or later, Transformers 4.57.0 or later and llama.cpp b6907 or later