---
type: "post"
id: "liquid-ai-released-lfm2-5-24b-a2b-a-c47d5c"
slug: "liquid-ai-released-lfm2-5-24b-a2b-a-c47d5c"
title: "Liquid AI released LFM2.5-24B-A2B , a 24B MoE with only 2.45B active parameters per token."
description: "**Thesis:** A model release becomes practically testable when its published artifacts and launch path are explicit."
retrieval_nugget: "**Thesis:** A model release becomes practically testable when its published artifacts and launch path are explicit."
published_at: "2026-10-18"
updated_at: "2026-10-18"
record_date: "2026-10-18"
date_kind: "published_at"
topics: ["models","open-models","inference"]
source_urls: ["https://huggingface.co/LiquidAI/LFM2.5-24B-A2B-GGUF"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:liquid-ai-released-lfm2-5-24b-a2b-a-c47d5c"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:00d3cce9844032183e4e3fae8b22a965dc6d09956ca8b34b27f7ce0f8240ad81","reviewed_at":"2026-10-18T23:01:09.559Z","source_evidence_count":1,"verified_claim_count":2,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":2,"implication_count":1,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"A model release becomes practically testable when its published artifacts and launch path are explicit.","observed_facts":[{"text":"The repository lists BF16, F16, Q4_K_M, Q5_K_M, Q6_K, Q8_0, and QAD-Q4_0 GGUF files.","source_urls":["https://huggingface.co/LiquidAI/LFM2.5-24B-A2B-GGUF"]},{"text":"The repository provides llama.cpp instructions for serving and terminal inference and lists additional local tools.","source_urls":["https://huggingface.co/LiquidAI/LFM2.5-24B-A2B-GGUF"]}],"mechanism":"Liquid AI’s Hugging Face repository pairs multiple GGUF weight variants with instructions for local inference, reducing the setup ambiguity for developers who want to evaluate the model in their own environment.","why_now":"The repository makes this release actionable for teams comparing local model options, while the available evidence remains limited to published files and setup guidance.","implications":["Builders can select a published artifact and begin workload-specific evaluation without treating the release announcement as a performance result."],"evidence_boundary":"The supplied repository excerpt documents artifact availability and instructions, but does not establish comparative quality, throughput, latency, or actual memory use across hardware.","watch_conditions":["Look for reproducible evaluations that identify the exact quantization, runtime, hardware, workload, and quality criteria."],"related_records":[],"new_branch_reason":"No prior New Runtime record was supplied, so this is a distinct analytical record rather than an asserted continuation of an earlier record."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/liquid-ai-released-lfm2-5-24b-a2b-a-c47d5c/","markdown":"https://newruntime.com/posts/liquid-ai-released-lfm2-5-24b-a2b-a-c47d5c.md","json":"https://newruntime.com/posts/liquid-ai-released-lfm2-5-24b-a2b-a-c47d5c.json"}
---

# Liquid AI released LFM2.5-24B-A2B , a 24B MoE with only 2.45B active parameters per token.

## Retrieval answer

**Thesis:** A model release becomes practically testable when its published artifacts and launch path are explicit.

**Thesis:** A model release becomes practically testable when its published artifacts and launch path are explicit. **Mechanism:** Liquid AI’s Hugging Face repository pairs multiple GGUF weight variants with instructions for local inference, reducing the setup ambiguity for developers who want to evaluate the model in their own environment. **Why now:** The repository makes this release actionable for teams comparing local model options, while the available evidence remains limited to published files and setup guidance.

The repository lists BF16, F16, Q4_K_M, Q5_K_M, Q6_K, Q8_0, and QAD-Q4_0 GGUF files. It also includes llama.cpp commands for serving and terminal inference, and lists additional local tools. **First implication:** Builders can select a published artifact and begin workload-specific evaluation without treating the release announcement as a performance result. **Evidence boundary:** The supplied repository excerpt documents artifact availability and instructions, but does not establish comparative quality, throughput, latency, or actual memory use across hardware.

**First watch condition:** Look for reproducible evaluations that identify the exact quantization, runtime, hardware, workload, and quality criteria. Until then, the file list is evidence of packaging choices, not proof that one variant is faster or better. This starts a distinct analytical record because no prior New Runtime record was supplied; the earlier-post context is not treated as a linked analytical record.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Models](https://newruntime.com/topics/models.md): Explore the models topic hub.
- topic hub: [Open Models](https://newruntime.com/topics/open-models.md): Explore the open-models topic hub.
- topic hub: [Inference](https://newruntime.com/topics/inference.md): Explore the inference topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)