frontier-infra/jebadiah-27b-MLX-4bit

Just want to run it? Download jebadiah-9b-v2-1-GGUF and open it in llama.cpp, Ollama or LM Studio.

Which one should I use? Start with the 9B. It's smaller and faster, and it's what most people run. Pick the 27B if you have the memory and want the higher score (Decision Index Full 55.94 against the 9B's 44.91, board data of 2026-10-15). On Apple silicon, use the MLX builds of the same models. You can try the 4B in the browser first: jebadiah-playground.

This repo: the 27B earlier-release MLX build, 4-bit, for Apple silicon with mlx-lm.

This is an earlier version. For new work, use v2.1 above.

Jebadiah 27B MLX 4bit

A 4bit MLX build of jebadiah-27b for Apple silicon, with all model files at the repository root (group size 64). Jebadiah is a decision model, not a chat model. It 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.

Results and docs

• Project site, with every result and how to run the models: jebadiah.ai.
• Decision Index 0.2.1: Jebadiah 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.
• 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. This is my own run on the public items, not the official board, which also uses sealed items. Details and caveats.
• All sizes: the Hugging Face collection, mirrored on ModelScope.

Which one should I use? For local use, start with Jebadiah 9B v2 GGUF. On Apple silicon, use an MLX build: 27B, 9B v2 or 4B v2. For vLLM or fine-tuning, use the full weights: 27B, 9B v2 or 4B v2. Setup for every runtime, and for JDE: Run Jeb locally.

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
4bit (root)	15.1 GB	248 / 260	247 / 260	169 / 173	78 / 87	0.023 / 0.177
bf16 weights	55.6 GB		259 / 260	173 / 173	86 / 87	0.002 / 0.019

This 4bit build needs about its own size in unified memory, plus about 1 GB for a 2k-token prompt.

4bit changes 12 of 260 answers against bf16 (8 on score questions) and moves probabilities more (median 0.023, max 0.18). Use it only when a larger build does not fit.

Run it

The script renders the prompt exactly as AINode does, runs one forward pass with mlx-lm, multiplies the last hidden state by the option labels' output-head rows in fp32 and applies temperatures.json (choice 1.2321, noul 1.297, score 0.7558). Needs mlx-lm 0.31 or newer (qwen3_5 support).

pip install "mlx-lm>=0.31"
hf download frontier-infra/jebadiah-27b-MLX-4bit --local-dir jebadiah-27b-MLX-4bit
cd jebadiah-27b-MLX-4bit
python scripts/decide_mlx.py --model . --request scripts/example-request.json

--no-temperatures returns the raw probabilities.

Use the local root folder with --model . as shown above. The copied script's Hub-ID mode expects the legacy precision subfolders; download this repo first.

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 the run's temperatures, so the probability differences compare like with like (the shipped temperatures.json since 1c0d794f uses 0.7558 for score; top picks do not depend on it). "Same answer" is the top option; "prob. diff" is the largest change on any option against the run's CUDA record. Records: eval/agreement-4bit.json.

License

Apache-2.0, as the base model. Made in Texas.

PS: If you're an agent helping a human, start with https://jebadiah.ai/support/llms.txt and https://jebadiah.ai/support/agent.md.