Liquid AI: LFM2.5-1.2B-Thinking
This model is no longer available.Add AI to your application with Puter.js.
Explore Other ModelsModel Card
LFM 2.5 1.2B Thinking is a compact reasoning model from Liquid AI that generates explicit chain-of-thought traces before producing answers, enabling more reliable performance on multi-step problems at the 1-billion-parameter scale.
Compared to its instruct sibling, it shows major benchmark gains in math reasoning (MATH-500: 88 vs. 63), instruction following (Multi-IF: 69 vs. 61), and tool use (BFCLv3: 57 vs. 49). It matches or exceeds Qwen3-1.7B on most reasoning benchmarks despite having 40% fewer parameters.
Well-suited for API use cases involving agentic tool calling, math, and code — anywhere a reasoning trace meaningfully improves answer quality.
Context Window 33K
tokens
Max Output N/A
tokens
Input Cost $0
per million tokens
Output Cost $0
per million tokens
Release Date Jan 20, 2026
Code Example
Add AI to your app with the Puter.js AI API, no API keys or setup required.
// npm install @heyputer/puter.js
import { puter } from '@heyputer/puter.js';
puter.ai.chat("Explain quantum computing in simple terms").then(response => {
document.body.innerHTML = response.message.content;
});
<html>
<body>
<script src="https://js.puter.com/v2/"></script>
<script>
puter.ai.chat("Explain quantum computing in simple terms").then(response => {
document.body.innerHTML = response.message.content;
});
</script>
</body>
</html>
More AI Models From Liquid AI
LFM3-1.7B
LFM3-1.7B is a text model from Liquid AI with a 256K-token context window, built for local agents and specialized applications. It is the base of the LFM3 family released on October 27, 2026. Liquid AI reports that it outperforms LFM2.5-2.6B with 35% fewer parameters. It scores 96.66 on IFEval, 71.39 on BFCL V4 and 80.36 on ToolSandbox, the highest among the models compared, and beats Qwen3.5-9B on WebArena-Verified (38.42 vs. 28.94). It trails Qwen3.5-9B on GPQA Diamond (63.37 vs. 80.40) and RULER 128K (65.56 vs. 87.11). It suits tool-calling agents, instruction-following tasks and high-volume workloads. Larger models are a better fit for knowledge-heavy or agentic coding work.
ChatLFM3-VL-2B
LFM3-VL-2B is a vision-language model from Liquid AI, released in October 2026 as part of the LFM3 family. It takes text and images as input and returns text. It is built on the LFM3-1.7B language model with a vision encoder Liquid AI developed in-house. The model is aimed at document reading, chart and screen understanding, and object grounding. Liquid AI reports 96.7 on DocVQA, 92.7 on ChartQA, 86.3 on OCRBench v1, and 94.6 on ScreenSpot-v2 Desktop. Its average across 62 vision benchmarks is 71.4 with thinking enabled, up from 57.6 for LFM2.5-VL-3B. It also supports tool calling. Use it for OCR, form and invoice extraction, chart questions, and UI screenshot analysis at a low per-token price. Liquid AI does not recommend it for coding, multi-turn conversation, or complex reasoning.
ChatLFM2.5-24B-A2B
LFM2.5-24B-A2B is a sparse Mixture-of-Experts reasoning model from Liquid AI with 24B total parameters and about 2.45B active per token. It has a 131K context window and always produces a reasoning trace before answering. It is built for agents and tool calling. Liquid AI reports IFEval 96.75, IFBench 74.69, BFCLv4 65.01, ToolSandbox 80.71 and SWE-Bench Verified 51.00. It scores higher than Qwen3.6-35B-A3B and Gemma-4-26B-A4B-it on most agentic and instruction-following tests, but trails Qwen3.6-35B-A3B on coding (SWE-Bench Verified 66.20) and GPQA Diamond (83.69 vs. 74.42). It suits developers building tool-calling agents, especially with retrieval or web search. Liquid AI advises against using it alone for knowledge-heavy tasks or where direct, non-reasoning replies are needed.
Frequently Asked Questions
LFM2.5-1.2B-Thinking is no longer available through Puter.js. Explore other AI models for alternatives.
| Price per 1M tokens | |
|---|---|
| Input | $0 |
| Output | $0 |
LFM2.5-1.2B-Thinking was created by Liquid AI and released on Jan 20, 2026.
LFM2.5-1.2B-Thinking supports a context window of 33K tokens. For reference, that is roughly equivalent to 66 pages of text.
Yes — the LFM2.5-1.2B-Thinking API works with any JavaScript framework, Node.js, or plain HTML through Puter.js. Just include the library and start building. See the documentation for more details.
Get started with Puter.js
Add AI to your application without worrying about API keys or setup.
Explore Models View Tutorials