1.7B
- input
- $0.35
- cached
- $0.04
- output
- $0.50
- Context
- 128Ktokens
- Listed
- 3mo ago
- Size
- 2.6Bparameters
Cost/1M tokens
Liquid AI's LFM2.5-2.6B is post-trained from LFM2.5-2.6B-Base and belongs to the LFM2.5 family, which builds on the LFM2 architecture for on-device deployment. Its launch announcement centers on running agents on-device: the maker says the model is small enough for a phone, fast enough on CPUs, and lets agents run without cloud APIs. It is built for agentic work such as tool use and multi-step tasks, though its model card advises against relying on it for agentic coding and knowledge-heavy tasks.
- Language model
- Open weights · 2.6B
- Reasoning
- Tools
- Structured output
How it compares
Size class
Tiny2.6BAn open-weight model under 5 billion parameters, small enough for phones, laptops and edge devices.
Open models of a similar size from other labs
Vs. your stack
Add the models you use to see how this one compares.
Sign inFrequently asked questions
Running locally
Estimated 2.3 GB at Q4, 3.7 GB at Q8 and 6.5 GB at FP16. These are estimates, covering weights plus about 15% and 0.5 GB for a short context.
Yes, it fits at FP16 (estimated 6.5 GB) on 24 GB, 16 GB, 12 GB and 8 GB GPUs.
Yes, at Q4 it needs an estimated 2.3 GB. Assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits Q4 is an estimated 8 GB.
The smallest published build at 4-bit or higher is lfm2.5:latest (Q4_K_M), a 5.2 GB download. Its VRAM use is estimated at 6.5 GB.
Run ollama run lfm2.5:latest. That build is a 5.2 GB download and there are 6 other local Ollama builds.
The only runtime with recorded support is Ollama. No recorded support is listed for llama.cpp, vLLM, LM Studio or MLX.
The default Ollama build lfm2.5:latest is 5.2 GB. Of 7 builds with a published size, the smallest is 5.2 GB and the largest is 48 GB.
API and pricing
$0.35 per 1M input tokens and $0.50 per 1M output tokens. 1 paid provider is tracked.
Yes. Cached input costs $0.04 per 1M tokens versus $0.35 uncached, an 89% discount.
A request of 10,000 input tokens and 1,000 output tokens costs $0.004 at the cheapest provider, Liquid.
1 of 1 endpoint with uptime data last day was at 99% or better. Best uptime last day was 100.0%, by Liquid.
Privacy and data
Yes, the model has open weights. You can run it locally with ollama run lfm2.5:latest.
Capabilities and specs
The context window is 128K tokens (128,000) and the maximum output is 8K tokens (8,192).
No, only text. The listed input modality is text, with no image, audio, video or PDF support.
Yes, tool calling (function calling) is supported.
Yes, structured output (JSON schema or JSON mode) is supported.
Yes, a reasoning mode is supported and it is always on. 1 of 1 tracked providers expose reasoning, and no provider offers a reasoning_effort setting.
Quality and benchmarks
Matched on model kind, blended price and release date: Muse Glimmer 30B by Meta at $0.525 per 1M tokens, GLM 5.3 by Z.ai at $0.4024 per 1M tokens and Motif 3 by Motif Technologies at $0.525 per 1M tokens.
Muse Glimmer 30BLFM2.5-2.6B vs Muse Glimmer 30BGLM 5.3LFM2.5-2.6B vs GLM 5.3Motif 3LFM2.5-2.6B vs Motif 3
Versions and lineage
Both have a 128K token (128,000) context. LFM2.5-2.6B costs $0.35 input and $0.50 output per 1M tokens, LFM2.5-24B-A2B costs $0.05 input and $0.70 output. LFM2.5-2.6B was released August 11, 2026, LFM2.5-24B-A2B October 13, 2026.
Licensing and access
LFM2.5-2.6B has open weights.
The weights are on the Hugging Face repository LiquidAI/LFM2.5-2.6B. There are also 7 Ollama builds.
Basics
LFM2.5-2.6B was made by Liquid AI. It was first listed on OpenRouter on August 11, 2026.
No. The listed API price is $0.35 input and $0.50 output per 1M tokens. The model has open weights.
Generated from Vertical Tabs data, updated November 11, 2026.
Discussion
No comments yet
Be the first to share your thoughts.