Versions and compatibility
A released profile is immutable. Changing the model, revision, precision, runtime, flags, prompt set, token shape, assets, step count, seeds or metric definitions is a major version bump that creates a separate leaderboard.
Current profile versions
| Profile | Version | State | Model | Runtime | Profile hash |
|---|---|---|---|---|---|
| diffusion-enthusiast-base | 0.1.0 | Release candidate | stabilityai/stable-diffusion-3.5-large | diffusers 0.39.0 | 142edc9e59f5b3… |
| diffusion-entry-base | 0.1.0 | Release candidate | stabilityai/stable-diffusion-xl-base-1.0 | diffusers 0.39.0 | d1eed7cc6558ed… |
| diffusion-frontier-base | 0.1.0 | Release candidate | black-forest-labs/FLUX.1-schnell | diffusers 0.39.0 | e35d8f0880ddcb… |
| llm-enthusiast-base | 0.1.0 | Release candidate | Qwen/Qwen3-8B | vllm 0.26.0 | 61ec8da4e3a2e0… |
| llm-enthusiast-int4 | 0.1.0 | Release candidate | Qwen/Qwen3-32B-AWQ | vllm 0.26.0 | 4acbd6459badce… |
| llm-enthusiast-nvfp4 | 0.1.0 | Release candidate | nvidia/Qwen3-32B-NVFP4 | vllm 0.26.0 | 4643cf31dded44… |
| llm-entry-base | 0.1.0 | Release candidate | Qwen/Qwen3-4B-Instruct | vllm 0.26.0 | c976c1bd573623… |
| llm-entry-int4 | 0.1.0 | Release candidate | Qwen/Qwen3-8B-AWQ | vllm 0.26.0 | 786cba1eae562a… |
| llm-entry-nvfp4 | 0.1.0 | Release candidate | nvidia/Qwen3-8B-NVFP4 | vllm 0.26.0 | 66f58d0e43ecc1… |
| llm-frontier-base | 0.1.0 | Release candidate | Qwen/Qwen3-32B | vllm 0.26.0 | 98cfb0979cbf81… |
| llm-frontier-int4 | 0.1.0 | Release candidate | Qwen/Qwen2.5-72B-Instruct-AWQ | vllm 0.26.0 | 5ac4055c2a6495… |
| llm-frontier-nvfp4 | 0.1.0 | Release candidate | nvidia/Qwen2.5-72B-Instruct-NVFP4 | vllm 0.26.0 | 8c27b806d4f34a… |
| vision-enthusiast-base | 0.1.0 | Release candidate | Qwen/Qwen3-VL-8B-Instruct | vllm 0.26.0 | c5243aa080b5c3… |
| vision-entry-base | 0.1.0 | Release candidate | Qwen/Qwen3-VL-4B-Instruct | vllm 0.26.0 | 70b66e37e396f7… |
| vision-frontier-base | 0.1.0 | Release candidate | Qwen/Qwen3-VL-32B-Instruct | vllm 0.26.0 | 405cabdb56c02c… |
Version policy
Release candidates
Profiles at version 0.x are not frozen. Their model revision, runtime version and harness commit are still being pinned against the tier's minimum-VRAM reference system, and their leaderboards are provisional. Results collected against a 0.x profile are ranked only against that same version, and are not migrated when it freezes.
Freezing a profile requires conformance data from at least two GPU classes including the tier minimum, a repeatability coefficient of variation at or under 3%, and a completed licence review of every model and asset it references.
Container images
Every released image is multi-architecture (linux/amd64 andlinux/arm64), signed with Cosign, and published with an SBOM and build provenance. A result reaches Verified only if its recorded image digest matches a released one.
ghcr.io/pavel4ai/localmax-llm ghcr.io/pavel4ai/localmax-vision ghcr.io/pavel4ai/localmax-diffusion
Released digests are tracked incontainers/released-images.jsonand compiled into the validator, so the accepted list is itself public and versioned.