LFM2.5-2.6B

Liquid AI

LFM2.5-2.6B is Liquid AI's open-weight 2.6B dense hybrid text→text model (LFM2 architecture) post-trained for on-device agentic workloads: planning, native tool calling, and multi-step tasks. ~2.69B parameters, trained on ~34T tokens, 128K context (131,072 tokens), 128K vocabulary. Always thinks before answering. Day-one support for Transformers, vLLM, SGLang, llama.cpp, MLX, and ONNX. Hugging Face: LiquidAI/LFM2.5-2.6B (+ Base). License: LFM Open License v1.0 (lfm1.0).

Key Specifications

Parameters
2.7B
Context
-
Release Date
August 4, 2026
Average Score
55.5%

Timeline

Key dates in the model's history
Announcement
August 4, 2026
Last Update
September 10, 2026
Today
September 19, 2026

Technical Specifications

Parameters
2.7B
Training Tokens
34.0T tokens
Knowledge Cutoff
-
Family
-
Capabilities
MultimodalZeroEval

Benchmark Results

Model performance metrics across various tests and benchmarks

Other Tests

Specialized benchmarks
AIME 2025
Vendor table (AIME25)Self-reported
51.9%
BFCL-V4
Vendor table (BFCLv4); temperature=0.001, max output 4096Self-reported
56.9%
Claw-Eval
Vendor table; Claw-Eval average (EN)Self-reported
62.8%
IFBench
Vendor tableSelf-reported
59.2%
LiveCodeBench v6
Vendor table (LiveCodeBenchv6)Self-reported
59.4%
Multi-IF
Vendor tableSelf-reported
80.1%
PinchBench
Vendor table; temperature=0.6, max output 8192Self-reported
68.2%
Tau3 Banking
Liquid harness (vendor table τ³-Bench Banking); temperature=0, no output limit. Scale differs from some AA-reported τ³-Banking rows.Self-reported
5.7%

License & Metadata

License
lfm1_0
Announcement Date
August 4, 2026
Last Updated
September 10, 2026

Compare LFM2.5-2.6B

All comparisons

Similar Models

All Models

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.