LBS-27B vs NVIDIA Nemotron 3.5 Bolt E4B 4GB

LBS-27B

A 27B Qwen3.8 fine-tune for GUI grounding that zooms into screenshots over a few turns before naming a point, and inclusionAI says it also helps with fine detail in large images. It needs about 96 GB at BF16 and a client that runs its crop actions. Every result so far is inclusionAI's own, with no safety results or injection guidance.

NVIDIA Nemotron 3.5 Bolt E4B 4GB

A small cousin for voice and device assistants, game characters and robot interfaces, trained for multi-turn tool calls, instruction hierarchy and short spoken replies, with thinking switchable per request. NVIDIA's Q4_K_M GGUF needs about 8 GB, and Ollama and LM Studio carry it. Context is 64K, half the Gemma 4 base, and NVIDIA publishes no safety results.

LBS-27B

1. Serve with vllm serve inclusionAI/LBS-27B --tensor-parallel-size 4 --max-model-len 65536, which is what the card's wrapper script runs; keep the context at 32,768 or more for multi-turn images
2. Build the system prompt with build_system_prompt from the LBS branch's zoomin_grounding_protocol.py, send the screenshot after image_path: <name>, and don't pass tools; the model answers in plain text
3. Parse <action>crop_image(image_path=..., x1=..., y1=..., x2=..., y2=...)</action>, cut the crop yourself and return it with its mapping to the original image; allow at most two crops per turn and three per episode
4. Read the final answer as [x.xx,y.yy] on a 0 to 1000 scale of the original image, and use your own executor to click
5. Pass reasoning_effort as xhigh, medium or low only, since the template raises an error for high, and don't count on thinking_budget, which the template ignores
6. Don't turn on MTP speculative decoding; the MTP weights aren't in the repository though config.json names one MTP layer

NVIDIA Nemotron 3.5 Bolt E4B 4GB

1. Quickest start is ollama run nemotron-3.5-bolt:e4b, about 8 GB of files, and set num_ctx to 65536 or less, since NVIDIA supports 64K
2. With NVIDIA's GGUF use llama-server -m Nemotron-3.5-Bolt-E4B-Q4_K_M.gguf -c 65536 -ngl 999 --jinja, and add --reasoning on --reasoning-budget <tokens> to cap thinking
3. With vLLM 0.31.0 or later pass --enable-auto-tool-choice --tool-call-parser gemma4 --reasoning-parser gemma4 and the repository's chat_template.jinja
4. Pass enable_thinking explicitly in chat_template_kwargs, since the template defaults to off whatever the card says, and set skip_special_tokens false as the card's examples do
5. Use temperature=1.0 and top_p=0.95 in both modes, as the card says
6. Set conversational_mode true for spoken replies, which appends a brevity instruction to the system prompt
7. Answer a slow tool at once with a pending tool message, then send the result later as a user message starting TOOL RESULT, as the card's asynchronous pattern says
8. Keep persistent rules in the system prompt and send state updates as separate user or system messages in order, per the card's prompting guidelines