Agents-A1-4B vs Gemma 4 31B IT

Agents-A1-4B

A small agent cousin for search, research and tool-calling loops, with vision and 262,144 tokens of context. It runs in about 4 GB from InternScience's Q4_K_M GGUF, in vLLM, SGLang, Transformers or llama.cpp. The main risks are an undeclared base, no safety results or agent guidance, and a template that adds its own dated system prompt when you send none.

Gemma 4 31B IT

A dense multimodal cousin with documented tool calling, optional thinking and a 256K context, for general agent work and image reading on one 24 GB GPU through Google's 4-bit build. It runs in every major runtime under plain Apache 2.0. The main risk is that Google publishes no safety results and has gone quiet on the repository since July.

Agents-A1-4B

1. Quickest start is llama-server -hf InternScience/Agents-A1-4B-Q4_K_M-GGUF --jinja, about 3.4 GB with the vision projector
2. With vLLM use --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder, as on the card, and add --language-model-only for text-only work
3. Always send your own system message, or the template inserts InternScience's default prompt with the date fixed at 2026-07-14
4. The card recommends temperature 0.85, top_p 0.95, top_k 20, min_p 0 and presence_penalty 1.1, and there is no generation_config.json to set them
5. Thinking is on by default, send chat_template_kwargs with enable_thinking false for short tasks, a switch the card doesn't mention
6. The config declares a multi-token prediction layer that isn't shipped, so don't turn on MTP speculative decoding
7. Validate JSON output yourself, since structured output isn't documented

Gemma 4 31B IT

1. On the reference machine run llama-server -hf google/gemma-4-31B-it-qat-q4_0-gguf --hf-file gemma-4-31B_q4_0-it.gguf --jinja -c 32768, adding gemma-4-31B-it-mmproj.gguf for images
2. Or ollama run gemma4:31b for the 20 GB Q4_K_M build
3. In vLLM serve with --enable-auto-tool-choice --tool-call-parser gemma4 --reasoning-parser gemma4, as vLLM's recipe does
4. Thinking is off by default, so pass enable_thinking: true in chat_template_kwargs to turn it on, and expect the reasoning in the thought channel block
5. Drop earlier thoughts from the history except on tool-call turns, as the card asks
6. Use the card's sampling, temperature 1.0, top_p 0.95, top_k 64
7. Validate function names and arguments before running a call, and check JSON output yourself or use the runtime's guided decoding
8. The w4a16 build relies on Marlin INT4 kernels that vLLM marks unsupported on AMD, so use the GGUF on the reference machine