Home/Events/Launch of LFM2.5-24B-A2B by LiquidAI for On-Device AI Agents

Launch of LFM2.5-24B-A2B by LiquidAI for On-Device AI Agents

Confirmed

Confidence

90%

Impact: 80%

Updated 5h ago

Consensus Brief

LiquidAI has introduced LFM2.5-24B-A2B, a 24 billion parameter reasoning model designed for efficient on-device AI agents. The model operates with only 2 billion active parameters per token, allowing for cost-effective reasoning and high inference speeds on various hardware.

Sourced from

Hugging FaceT1

Primary: Hugging Face

What Changed Since Last Update

5h ago

LFM2.5-24B-A2B features a new architecture that enhances reasoning capabilities and increases the context window from 32K to 131K tokens, along with a significant increase in pre-training data from ~17 trillion to ~33 trillion tokens.

Claim Ledger

4 claims tracked across sources

Confirmed Fact

LFM2.5-24B-A2B achieves 184 tokens/s on an Apple M5 Max and up to 21K output tokens/s on a single NVIDIA B300 GPU.

Confirmed Fact

LFM2.5-24B-A2B is competitive with Qwen3.6-35B-A3B, scoring 48.09 on LiveClaw versus 38.19.

Confirmed Fact

The model has a context window that grows from 32K to 131K tokens.

Confirmed Fact

LFM2.5-24B-A2B leads or matches the competition on five of the eight tool-use and agentic benchmarks.

Role-Based Impact Analysis

Source Timeline

1 source corroborating

Run fast, private agents on-device with LFM2.5-24B-A2B

Hugging Face·5h ago