--- license: apache-2.0 base_model: frontier-infra/jebadiah-4b-v2 base_model_relation: quantized library_name: gguf language: - en tags: - gguf - llama.cpp - decision-model - system-one - typed-decisions - calibrated-probabilities - ainode --- # Jebadiah 4B v2 GGUF GGUF builds of [Jebadiah 4B v2](https://huggingface.co/frontier-infra/jebadiah-4b-v2) for [llama.cpp](https://github.com/ggml-org/llama.cpp), which runs on NVIDIA, AMD and Apple GPUs and on plain CPUs. Jebadiah answers a typed question (choice, noul or score) with a probability for every option, read from one forward pass. Nothing is generated. Code, trainer and evals: [getainode/jebadiah](https://github.com/getainode/jebadiah). **Results and docs** - Project site, with every result and how to run the models: [jebadiah.ai](https://jebadiah.ai). - For the family: [Decision Index 0.2.1](https://huggingface.co/spaces/multimodalart/jev-decision-index): the 27B scores 54.67, #5 of 67 open models (as of 2026-09-26). This is the board's own number; the maintainer validated my run and put it on the leaderboard. [Run record](https://huggingface.co/datasets/frontier-infra/jebadiah-decision-index-results/tree/main/runs/jebadiah-27b-1c0d794f). - JevBench v1.4.2: on its 231 public items, run through its own harness, Jebadiah 27B scores 0.866, the same as Jev 1.13.0; Jebadiah 9B v2 scores 0.818. This is my own run on the public items, not the official board, which also uses sealed items. [Details and caveats](https://github.com/getainode/jebadiah#where-we-stand-on-jevbench). The 4B has not been run on either. - All sizes: the [Hugging Face collection](https://huggingface.co/collections/frontier-infra/jebadiah-open-system-one-decision-models-6ab80765ddd3fa0b3eba5213), mirrored on [ModelScope](https://www.modelscope.ai/profile/JasonBrashear). **Which one should I use?** For local use, start with [Jebadiah 9B v2 GGUF](https://huggingface.co/frontier-infra/jebadiah-9b-v2-GGUF). On Apple silicon, use an MLX build: [27B](https://huggingface.co/frontier-infra/jebadiah-27b-MLX), [9B v2](https://huggingface.co/frontier-infra/jebadiah-9b-v2-MLX) or [4B v2](https://huggingface.co/frontier-infra/jebadiah-4b-v2-MLX). For vLLM or fine-tuning, use the full weights: [27B](https://huggingface.co/frontier-infra/jebadiah-27b), [9B v2](https://huggingface.co/frontier-infra/jebadiah-9b-v2) or [4B v2](https://huggingface.co/frontier-infra/jebadiah-4b-v2). ## Files Every file was checked on the 260 held-out questions the merged weights were checked on, and compared with the merged bf16 weights and with the training run's own eval records. | File | Size | Same answer as bf16 | Same as the run | choice + noul | score | Prob. diff median / max | |---|---:|---:|---:|---:|---:|---:| | `jebadiah-4b-v2-Q8_0.gguf` | 4.6 GB | **256 / 260** | 257 / 260 | 171 / 173 | 86 / 87 | 0.004 / 0.053 | | `jebadiah-4b-v2-Q5_K_M.gguf` | 3.2 GB | **239 / 260** | 240 / 260 | 166 / 173 | 74 / 87 | 0.016 / 0.119 | | `jebadiah-4b-v2-Q4_K_M.gguf` | 2.8 GB | **236 / 260** | 237 / 260 | 164 / 173 | 73 / 87 | 0.022 / 0.355 | | *bf16 weights* | | | 259 / 260 | 172 / 173 | 87 / 87 | 0.002 / 0.019 | Which one: `Q8_0` if it fits (it changed 4 answers here); `Q4_K_M` when memory is short. A file needs about its own size in GPU or unified memory, plus about 1 GB for a 4k context. **`Q5_K_M` changes 21 of 260 answers** against bf16 (13 on score questions) and moves probabilities more (median 0.016, max 0.12). Use it only when a larger build does not fit. **`Q4_K_M` changes 24 of 260 answers** against bf16 (14 on score questions) and moves probabilities more (median 0.022, max 0.36). Use it only when a larger build does not fit. ## Run it The answer is the log probability of each option label ("A", "B", ...) at the answer position, which llama-server's `/completion` returns. The script renders the prompt exactly as AINode does, sends the raw text (so the server's own chat template is never used), renormalises over the labels and applies `temperatures.json` (choice 1.1167, noul 1.3319, score 1.1974). You need a llama.cpp that knows the `qwen35` architecture: we checked v0.5.0 (older builds refuse the file). ```bash hf download frontier-infra/jebadiah-4b-v2-GGUF --include "*Q8_0.gguf" "scripts/*" "*.json" "*.jinja" "*.txt" --local-dir jebadiah-4b-v2-GGUF cd jebadiah-4b-v2-GGUF llama-server -m jebadiah-4b-v2-Q8_0.gguf -c 4096 -np 1 --port 8080 pip install transformers # the tokenizer only, no torch python scripts/decide_gguf.py --server http://127.0.0.1:8080 --request scripts/example-request.json ``` `--no-temperatures` returns the raw probabilities. For LM Studio, see the next section. Ollama was not checked: a decision needs the log probability of every option label at one position; if your runtime cannot return those, use llama-server. On `example-request.json` (jebadiah-4b-v2-Q8_0.gguf): ```json { "route": {"type": "choice", "choice": "billing", "confidence": 0.397366, "probabilities": {"billing": 0.598244, "support": 0.082792, "sales": 0.318963}}, "urgent": {"type": "noul", "noul": 0.153519} } ``` ## Use it in LM Studio Jeb works in LM Studio through its local server, not the chat window: chat runs with thinking on and shows text, while a decision needs the probability of every option label. `scripts/decide_lmstudio.py` takes the same arguments and prints the same output as `decide_gguf.py`. It sends AINode's messages to LM Studio's `/v1/chat/completions` with thinking off (`"reasoning_effort": "none"`), where LM Studio renders the same prompt text the llama-server path sends, and reads the option labels from the top log probabilities that come back. It stops with an error if LM Studio's prompt length differs from the local tokenizer's. 1. In LM Studio, search for `jebadiah-4b-v2` and download `jebadiah-4b-v2-Q8_0.gguf` from this repository. 2. Open the **Developer** tab, start the server and load the model. Note the identifier LM Studio shows for it (for example `jebadiah-4b-v2`). 3. In a terminal: ```bash hf download frontier-infra/jebadiah-4b-v2-GGUF --include "scripts/*" "*.json" "*.jinja" "*.txt" --local-dir jebadiah-4b-v2-GGUF pip install transformers # the tokenizer only, no torch python jebadiah-4b-v2-GGUF/scripts/decide_lmstudio.py --model jebadiah-4b-v2 --request jebadiah-4b-v2-GGUF/scripts/example-request.json ``` If **Require Authentication** is on in LM Studio's server settings, create a token there and `export LM_API_TOKEN=...` first. Tested on the 9B only: LM Studio 0.4.21 with `jebadiah-9b-v2-Q8_0`, 257 of 260 answers the same as bf16, and the same answer as llama-server on the same file on 260 of 260. This 4B build uses the same script and the same prompt, but it has not been run in LM Studio. **At most 20 options per question.** LM Studio returns only the top 20 log probabilities, the same cap AINode's own route has. On the 77-option Banking77 questions the pick was still right, but the probabilities moved by up to 0.16, so do not rely on them past 20 options. **If LM Studio shows a "Vision" tag, ignore it.** Some GGUF repositories, including third-party quants of Jeb, ship a vision file (mmproj) from the Qwen base, and LM Studio labels the model Vision because of it. Jeb was not trained on images, and it does not make decisions in the chat window. It decides only through `decide_gguf.py` or `decide_lmstudio.py`, which send the typed question with thinking off and read the probabilities. The mmproj is not needed. ## How it was measured Jevals PubMedQA, Banking77 (77 options) and HelpSteer2, plus Nimble: the merge check's fixed sample (seed 20260925), the run's option order and temperatures. "Same answer" is the top option; "prob. diff" is the largest change on any option against the run's CUDA record. llama-server ran on Metal (M3 Ultra) with the same tokens as the Python renderer on every prompt. Records: `eval/agreement-*.json`. ## License Apache-2.0, as the base model. Made in Texas.