LFM2.5-1.2B-Thinking is a lightweight reasoning-focused model optimized for agentic tasks, data extraction, and RAG—while still running comfortably on edge devices. It supports long context (up to 32K tokens) and is...
- Modalities
- text → text
- In / out per 1M
- Free / Free
- Context
- 33K tokens
- Added
- Jan 20, 2026
- Tokenizer
- Other
- Reasoning
- always on
Pricing
Per 1M tokens. The provider price and our flat 3% fee are separate columns — what you pay is their sum.
| Per 1M tokens | Provider | + 3% fee | You pay |
|---|---|---|---|
| Input | Free | — | Free |
| Output | Free | — | Free |
Providers
No provider breakdown is published for LFM2.5-1.2B-Thinking (free) in the current catalog snapshot (Jul 28, 2026). Browse all providers
Supported parameters
- frequency_penalty
- include_reasoning
- max_tokens
- min_p
- presence_penalty
- reasoning
- repetition_penalty
- seed
- stop
- structured_outputs
- temperature
- tool_choice
- tools
- top_k
- top_p
Call it
OpenAI-compatible: point your SDK at api.openkey.ai/v1 and use
model liquid/lfm-2.5-1.2b-thinking:free.
curl https://api.openkey.ai/v1/chat/completions \
-H "Authorization: Bearer ***" \
-H "Content-Type: application/json" \
-d '{
"model": "liquid/lfm-2.5-1.2b-thinking:free",
"messages": [{"role": "user", "content": "Hello"}]
}'import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.openkey.ai/v1",
api_key=os.environ["OPENKEY_API_KEY"],
)
completion = client.chat.completions.create(
model="liquid/lfm-2.5-1.2b-thinking:free",
messages=[{"role": "user", "content": "Hello"}],
)
print(completion.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.openkey.ai/v1",
apiKey: process.env.OPENKEY_API_KEY,
});
const completion = await client.chat.completions.create({
model: "liquid/lfm-2.5-1.2b-thinking:free",
messages: [{ role: "user", content: "Hello" }],
});
console.log(completion.choices[0].message.content);About LFM2.5-1.2B-Thinking (free)
LFM2.5-1.2B-Thinking is a 1.2B-parameter reasoning model from Liquid AI built on liquid neural network architecture rather than a standard transformer stack. Reasoning is mandatory on every call, so it always works through a chain of thought before answering. With a 32,768-token context window and tool calling, structured outputs, and vision-free text-in/text-out support, it targets agentic tasks, data extraction, and RAG pipelines where you need a small model that still reasons, and it's light enough to run on edge devices.
This model's 32,768-token context window is larger than only 5% of models in the OpenKey catalog — it's built for small, fast footprints, not long documents. Reasoning is always on (not optional), which sets it apart from its sibling LFM2.5-1.2B-Instruct (free), the non-reasoning version at the same 1.2B size and same $0 price. Both are free via OpenKey, while the larger LFM2-24B-A2B sibling runs $0.03/M input and $0.12/M output.
Questions
- How much does LFM2.5-1.2B-Thinking cost via API?
- It's free. Provider pricing is $0 per 1M input tokens and $0 per 1M output tokens, and OpenKey's 3% fee on a $0 list price still comes out to $0. You get the same reasoning-focused 1.2B model with no per-token charge.
- Is LFM2.5-1.2B-Thinking free to use?
- Yes, it's one of OpenKey's 25 free models, with both prompt and completion tokens priced at $0/M from the provider. It's listed under the `:free` model ID, distinguishing it from any metered variant Liquid AI might release later.
- What is LFM2.5-1.2B-Thinking's context window?
- The context window is 32,768 tokens (32K), enough for roughly 20-25 pages of text in a single request. That's larger than only 5% of models in the OpenKey catalog, so it's built for short-to-medium agentic and RAG tasks, not long-document analysis.
- Does LFM2.5-1.2B-Thinking support tool calling and structured outputs?
- Yes. Its supported parameters include `tools`, `tool_choice`, and `structured_outputs`, along with `reasoning`/`include_reasoning` controls, `temperature`, `top_p`, `top_k`, and standard sampling penalties. It's text-in/text-out only — no image or audio modalities.
- How does LFM2.5-1.2B-Thinking compare to LFM2.5-1.2B-Instruct?
- Both are 1.2B-parameter models from Liquid AI priced at $0/M for input and output — the difference is reasoning. LFM2.5-1.2B-Thinking always runs a reasoning pass before answering (mandatory), while LFM2.5-1.2B-Instruct (free) is the direct-response version without that forced chain-of-thought step.
More from Liquid AI
All Liquid AI models →
Liquid AILFM2-24B-A2B
LFM2-24B-A2B is the largest model in the LFM2 family of hybrid architectures designed for efficient on-device deployment. Built as a 24B parameter Mixture-of-Experts model with only 2B active parameters per...
ctx 128K in $0.031 out $0.124
Liquid AIFreeLFM2.5-1.2B-Instruct (free)
LFM2.5-1.2B-Instruct is a compact, high-performance instruction-tuned model built for fast on-device AI. It delivers strong chat quality in a 1.2B parameter footprint, with efficient edge inference and broad runtime support.
ctx 33K in Free out Free