LFM2.5 24B A2B Q4_K_M (llama.cpp)

Metrics

Checklist score

– / 100

Design (rubric, 0–5)

–

Stop reason

done

Model time

2mins

Tool time (wall minus model)

2s

Wall time

2mins

Tool calls

–

Tool errors

–

Errors by type

–

Loop-guard events

–

Tokens in / out

427,848.0 / 27,674.0

Thinking tokens

–

Decode / prefill tok/s

199.2 / 3,598.7

Prefix reuse

0.983

Peak VRAM (GPU / engine)

15,081 / 15,072 MiB

Peak engine RAM

1,846 MiB

Engine

llamacpp

Model

LFM2.5 24B A2B Q4_K_M (llama.cpp) (lfm25-24b-a2b)

Settings

{"thinking": "medium", "reasoning_budget_tokens": 3072, "max_tokens": 7168, "sampling": "card: temp 0.1 top_k 50 repeat_penalty 1.05 (server)", "note": "harness model id is a Strata alias name; llama-server serves its one model whatever the request names"}

Node

omarchy (RTX 3090 24 GB)

Prompt

scene-pagoda-v1

Started

2026-10-30T01:17:54Z

Feature scores