Liquid AI launches a 2.6-billion-parameter on-device model that outperforms Qwen3.5-9B in three tests.

iconChainthink
Share
AI summary iconSummary
On August 5, Liquid AI released its on-device model LFM2.5-2.6B, featuring 2.6 billion parameters, marking a significant update in AI + crypto news. The model is open-sourced and consumes approximately 2.5GB of memory after quantization, achieving 30 tokens per second on mobile devices. In three benchmarks—tool calling, multi-turn instruction following, and structured output—it outperformed Qwen3.5-9B. Although Qwen3.5-9B, with 9.7 billion parameters, remains superior in code generation and complex agent tasks, this on-chain development underscores the ongoing competition in on-device AI performance.

ChainThink reports that on August 5, according to official news from Liquid AI, Liquid AI has released the on-device Agent model LFM2.5-2.6B and opened its model weights.

The model has 2.6 billion parameters, occupies approximately 2.5 GB of memory after quantization, and generates text at a speed of about 30 tokens per second on mobile devices.

Official evaluations show that LFM2.5-2.6B outperforms Qwen3.5-9B in three tests: tool calling, multi-turn instruction following, and structured output.

Qwen3.5-9B has a parameter scale of 9.7 billion but still maintains leadership in code generation and certain complex agent tasks.

Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.