Recipe
lfm25-8b-a1b-q4km-rtx3070ti-llamacpp-tp1
lfm25-8b-a1b-q4km-rtx3070ti-llamacpp-tp1Candidate derived from lfm2-5-8b-a1b-q4-k-m-rtx-3060-ti-8gb-llama-cpp-tp1 for rtx-3070-ti-8gb; the launch contract is copied, the evidence is not. Pending acceptance on the target card.
Record
- Status
- validated
- Source
- 0xsero
- Engine
- llama.cpp
- Engine version
- b9653 / 9dbc6621a
- Accelerators
- 1
- Tensor parallel
- 1
- Context tokens
- 32,768
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/LiquidAI/LFM2.5-8B-A1B-GGUF- Repository
- LiquidAI/LFM2.5-8B-A1B-GGUF
- Status
- known
- Link type
- Exact Hub repository
Public Hugging Face repository confirmed by the Hub API.
Validated: pinned artifact, pinned runtime, and accepted evidence. This is a launch contract.
Docker configuration
llama.cpp
Container · validated · docker
Validated launch contract
- Image
ghcr.io/ggml-org/llama.cpp:server-cuda12-b10481@sha256:b2497f8834f5ecb4e38530f6bf2734b8e0be107ff48e4720145911c86930f2ce- Digest
sha256:b2497f8834f5ecb4e38530f6bf2734b8e0be107ff48e4720145911c86930f2ce- Port
- 8080
Launch arguments
-hfLiquidAI/LFM2.5-8B-A1B-GGUF--n-gpu-layers999--host0.0.0.0--port8080-c32768
| Flag | Value |
|---|---|
-hf | LiquidAI/LFM2.5-8B-A1B-GGUF |
--n-gpu-layers | 999 |
--host | 0.0.0.0 |
--port | 8080 |
-c | 32768 |
Environment
| Variable | Value |
|---|---|
LLAMA_CACHE | /root/.cache/huggingface |
Mounts
| Source | Target |
|---|---|
~/.cache/huggingface | /root/.cache/huggingface |
Launch
Exact materialization of this validated launch contract: digest-pinned image, pinned model revision, and the audited arguments. Self-contained — required assets are fetched from this registry and verified against their recorded sha256 before mounting. Also available as local-ai run lfm25-8b-a1b-q4km-rtx3070ti-llamacpp-tp1.
- Pulls the exact container image by sha256 digest — the bytes that were validated, not a floating tag.
- Fetches any required engine assets from this registry and verifies each against its recorded sha256; the audited launch script verifies them again inside the container before use.
- Downloads the pinned model revision into your Hugging Face cache on first run (reused afterwards).
- Serves an OpenAI-compatible API on
localhost:8080— point any client at it.
docker run --rm \
--gpus all \
--shm-size 16g \
-p 8080:8080 \
-e LLAMA_CACHE=/root/.cache/huggingface \
-v ~/.cache/huggingface:/root/.cache/huggingface \
--entrypoint /app/llama-server \
ghcr.io/ggml-org/llama.cpp:server-cuda12-b10481@sha256:b2497f8834f5ecb4e38530f6bf2734b8e0be107ff48e4720145911c86930f2ce \
-hf \
LiquidAI/LFM2.5-8B-A1B-GGUF \
--n-gpu-layers \
999 \
--host \
0.0.0.0 \
--port \
8080 \
-c \
32768Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 32,768 | — | 276.3 | 345.7 | accepted | lfm25-8b-a1b-q4km-rtx3070ti-llamacpp-tp1-acceptance |
Remaining fields
Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.
- hardware count
- 1
- hardware id
- rtx-3070-ti-8gb
- id
- lfm25-8b-a1b-q4km-rtx3070ti-llamacpp-tp1
- model instance id
- liquidai-lfm2-5-8b-a1b-gguf--q4-k-m
- recipe source
- 0xsero
- schema version
- local-ai-registry/v1
- speed sweep ids
- lfm25-8b-a1b-q4km-rtx3070ti-llamacpp-tp1-acceptance
- status
- validated
capabilities
engine
- name
- llama.cpp
- version
- b9653 / 9dbc6621a
serving
- max context tokens
- 32,768
- tensor parallel
- 1
Provenance & metadata (3)
facts
metadata
- acceptance · accepted at
- 2026-09-03T10:24:36Z
- acceptance · apis
- chat, messages, responses
- acceptance · harness
- validate_rented.py on vast RTX 3070 Ti
- acceptance · served model id
- LiquidAI/LFM2.5-8B-A1B-GGUF
- derived from
- lfm2-5-8b-a1b-q4-k-m-rtx-3060-ti-8gb-llama-cpp-tp1
provenance
- captured at
- 2026-09-02T16:44:02Z
sources
| captured at | kind | url |
|---|---|---|
| 2026-09-02T16:44:02Z | normalized-recipe | github.com/0xSero/local-ai-registry ↗ |