JEV-27B-VL vs joey-1

JEV-27B-VL

A cousin for typed decisions with probabilities over text and images, answering in one pass with optional thinking. It needs about 67 GB in BF16 on vLLM, with no AutoTrust 4-bit build or hosted API. The main risks are labels distilled from TypeSafe's Jev against its customer terms and serving through a repository script on a vLLM development build.

joey-1

A very small cousin for routing, triage and guardrail checks, answering several typed questions about one state, with text or images, as probabilities from one prefill. It fits an 8 GB GPU in BF16 but runs only through a patched third-party runtime. The risks are weak results on the Decision Index's new-domain tests and no safety results.

JEV-27B-VL

1. Quickest start is hf download autotrust/JEV-27B-VL then bash JEV-27B-VL/serve.sh, which serves POST /v1/decide on port 8000, but it runs the repository's serve_decide.py on a September 2026 vLLM development build
2. On a released vLLM use plain vllm serve with --enable-lora --lora-modules jev-decision=JEV-27B-VL/adapter_vllm and do the card's client-side arithmetic with decision_head.json and calibration.json
3. Keep --max-num-seqs 8, the card says larger batches return wrong System 1 probabilities on vLLM's multimodal LoRA path
4. When calling /v1/completions or /v1/chat/completions yourself pass top_k 0 and top_p 1.0, or generation_config.json's top_k 20 and top_p 0.95 truncate the probabilities
5. Choices beyond 16 options use labels the head never saw in training, and the response's adaptation field says wide-labels
6. Downscale images to about 448 px to keep vision tokens and latency down
7. Treat it as AutoTrust's model, not TypeSafe's, and its request shape differs from TypeSafe's /v1/systemone

joey-1

1. Start on a CUDA GPU in pytorch/pytorch:2.8.0-cuda12.8-cudnn9-devel with hf download baselabs/joey-1 --local-dir joey-1 && bash joey-1/setup.sh, which clones OneJev at 81ce62f, applies serve.patch and installs the Decision Index kit
2. Then call DecisionEngine("baselabs/joey-1", device="cuda:0").decide(SystemOneRequest(state=..., questions=...)) from qev, as the card shows
3. Don't load it with stock vLLM or Transformers. They ignore the block-bidirectional attention in config.json's modeling_overrides and give probabilities the model wasn't trained for
4. Keep each request under 16,384 tokens, the trained length, or the engine refuses it
5. Expect weaker answers on domains unlike its training mix, and threshold on the returned probability before acting
6. Without hardware, Baseten's /v1/systemone serves it as decision-models/baseten-joey-2b-latest with a Baseten key