LFM2.5-2.6B
LFM2.5-2.6B by Liquid AI is a device-native model built to deploy agents everywhere, running under 1GB on llama.cpp, MLX, ONNX, CoreML for edge inference.
Model Overview
Capabilities, design details, and architectural traits
LFM2.5-2.6B - Deploy Agents Everywhere
LFM2.5-2.6B is a device-native foundation model from Liquid AI, explicitly positioned to deploy agents everywhere across edge hardware. It targets the latency, privacy, and compute constraints of processors outside data centers - phones, laptops, cars, and embedded systems.
| Trait | Detail |
|---|---|
| Agent-first edge positioning | Announced under the title "Deploy Agents Everywhere" - agentic deployment is the primary design goal |
| Device-native footprint | Built for frontier AI on edge processors |
| Multi-runtime portability | Runs on llama.cpp, MLX, ONNX, CoreML, SGLang, and vLLM |
| Quantization-Aware Distillation | LFM2.5 Q4_0 technique for efficient 4-bit edge deployment |
| Final Token Preference Optimization | LFM2.5 family training method to reduce "doom loops" in agent outputs |
Edge-to-Cloud Ecosystem
The model belongs to Liquid AI's LFM2.5 family. Fine-tuning and deployment is handled through the LEAP SDK, which bakes models to target runtimes. Production partnerships with Mercedes-Benz and AMD embed LFM models in cars and processors.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Liquid AI | Anthropic | Anthropic |
| Release Date | August 4, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 131K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | Free Best Input Pricing | $4 | $2 |
| Output Pricing | Free Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | text | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 8.4 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | 7.7 | - | - |
| Agentic Index | 2.4 | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Liquid AI
Other models by Liquid AI
Top AI Models
Leading alternatives by intelligence score
