Laguna XS 2.1 vs Nemotron 3 Super 120B A12B

Laguna XS 2.1

A small coding and agent cousin with 3B active parameters, so it's quick on modest hardware. Poolside's 4-bit GGUF, also Ollama's default, needs about 24 GB, and every major runtime runs it. The main risks are that nobody at Poolside answers its discussions, there are no safety results, and the SGLang route runs repository code.

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.

Laguna XS 2.1

1. Quickest start is ollama run laguna-xs-2.1, 20 GB of files, the same Q4_K_M as Poolside's GGUF
2. On our reference machine use llama-server -hf poolside/Laguna-XS-2.1-GGUF:Q4_K_M --jinja -c 131072, from a llama.cpp release with the laguna architecture (b11772 has it), or BF16 with vLLM (66.9 GB)
3. With vLLM use --tool-call-parser poolside_v1 --reasoning-parser poolside_v1 --enable-auto-tool-choice, on a build with vllm-project/vllm#47311, or --tool-call-parser glm47 on older builds
4. Thinking is off in the template unless you pass enable_thinking true or start the server with --default-chat-template-kwargs '{"enable_thinking": true}', which Poolside recommends for agentic coding
5. Pass earlier reasoning back as reasoning_content in the history, since the card says reasoning may stop in later steps if it's dropped
6. Sampling per the card is temperature 1.0 and top_k 20
7. The card warns that Ollama on Apple Silicon may return empty output until llama.cpp#25389 ships
8. Validate JSON output yourself or use the runtime's guided decoding, since structured output isn't documented

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)