WUMBOLABS :: STATIC LAB LOCAL / EVIDENCE / OPEN
WUMBOLABS / EVALUATIONS CANONICAL MODEL EVALUATION

/evaluations/grug-12b

Grug 12B

Grug 12B — the current WumboLabs evidence state on one page: tested profiles, validated context, and the full chronological testing history. Each value is attributed to the profile and event that measured it.

/ latest evidence 2026-07-04

WumboLabs tests Grug 12B on real consumer hardware. This is the canonical model page: current state first, then every tested profile and every evidence event. Values are attributed to the profile and event that measured them; historical findings remain the evidence of their tested stack and are never silently replaced.

Current state

  • Recommended profile: llama.cpp Q4_K_M (grug-12b-llamacpp-q4km, current) — canonical evidence
  • Latest evidence: 2026-07-04 — LMX speed run

Tested profiles

llama.cpp Q4_K_M — CURRENT

Profile identity: grug-12b-llamacpp-q4km.

Runtime and artifact identity are described inside the event sections below (hand-authored records; no machine-readable export).

Status: current canonical/recommended tested surface.

Canonical Evidence Profile Metadata

Events on this profile:

Testing history

Newest first. Each event is one immutable testing/publication event; the exact scientific report lives in the canonical evidence repository linked at the top of each event.

2026-07-04 — LMX speed run

Benchmark Only — LMX local speed · profile: llama.cpp Q4_K_M · maturity: BENCHMARK_ONLY · status: BENCHMARK_ONLY (LMX local speed)

Canonical Evidence / Full Report View on GitHub

Identity and scope

  • Profile: grug-12b-llamacpp-q4km — llama.cpp Q4_K_M
  • Evidence maturity: BENCHMARK_ONLY
  • Evidence scope: performance
  • Hardware: WumboJetsII (RTX 5070 12GB)

LMX local speed evidence: 69.24 tok/s out (llama.cpp Q4_K_M).

This event is a bounded public-safe summary derived from retained WumboLabs evidence. It is not a WELP characterization, a universal model ranking, or a production-readiness proof. Values are attributed to the tested artifact, runtime, hardware, suite, and settings.

2026-07-04 — Honesty smoke + provenance refresh

Specialized Test — honesty/provenance · profile: llama.cpp Q4_K_M · maturity: SPECIALIZED_TEST · status: SPECIALIZED_TEST / UNSCORED_SMOKE

Canonical Evidence / Full Report View on GitHub

Identity and scope

  • Profile: grug-12b-llamacpp-q4km — llama.cpp Q4_K_M
  • Evidence maturity: SPECIALIZED_TEST
  • Evidence scope: specialized
  • Hardware: WumboJetsII (RTX 5070 12GB)

Honesty smoke (2026-07-04) and provenance-refresh re-run (2026-07-24) for the Grug practical record.

This event is a bounded public-safe summary derived from retained WumboLabs evidence. It is not a WELP characterization, a universal model ranking, or a production-readiness proof. Values are attributed to the tested artifact, runtime, hardware, suite, and settings.

Shared comparison events

This model appears in shared multi-model comparisons. Each report is stored once in WumboLabs/evaluations/shared-events/; the model's measured entries remain attributed to the same historical event:

Canonical evidence

All canonical public evidence lives in WumboLabs/evaluations. Each event links an immutable full-commit/path citation; each profile remains a distinct scientific identity, not a separate repository.

Legacy provenance