kev — local AI model by Jared Palmer (community)
Written by Jakub Rusinowski · Last updated
LoRA adapters plus a pointer head on Qwen3.5 bases: one prefill, many typed questions, calibrated with a stored temperature. One of the families llama.cpp's server answers natively at /v1/systemone, from official ggml-org GGUF builds. Apache-2.0.
Variants
The smallest kev variant needs about 4 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
How to run kev locally
Install Ollama, then pull the tag.
Served by llama-server (llama.cpp) (default port 9931). This model does not run in Ollama.
Pick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: kev 4B, kev 9BRecommended GPU
The cheapest catalogued GPU that runs kev locally (min 4 GB VRAM) is the Intel Arc B570 (10 GB).
kev — frequently asked questions
How much VRAM does kev need?
kev needs about 4 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: kev 4B (4 GB, Q4_K_M); kev 9B (7 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run kev on an RTX 4090 (24 GB)?
Yes — kev runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for kev?
Q4_K_M is the best balance of quality and VRAM for kev in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run kev locally?
kev is a decision model: it is called over an HTTP endpoint (/v1/systemone), not chatted with. Where a variant is available in Ollama 0.35 or newer, pull it with `ollama pull` and send requests to the local server; the others ship their own server. Each variant page shows the exact commands for that model.