Liquid AI

Liquid LFM2.5-2.6B

Name: Liquid LFM2.5-2.6B
Author: Liquid AI

Liquid LFM2.5-2.6B is a Liquid AI model with 131K context, Not verified in / Not verified out per million tokens, last verified October 25, 2026. Open weights: yes. Public-board scores are linked to the publisher; vendor-only figures stay in the watch-out, not on the board.

LFM2.5-2.6B is Liquid AI's downloadable hybrid small model for on-device and edge workflows; use its model-specific license and runtime requirements.

Recorded profile verification October 25, 2026Individual provider fields may change

AI model specification details
Specification	Liquid LFM2.5-2.6B
Provider	Liquid AI
Tier	Budget
Context window	131K
Max output	Not verifiedUnverified
Input / 1M tokens	Not verifiedUnverified
Output / 1M tokens	Not verifiedUnverified
Weights	Open
Parameters	2.6B
Reasoning levels	Not verifiedUnverified
Modalities	text
License	LFM Open License v1.0
Released	August 4, 2026

Pricing tiers: Downloadable weights under LFM Open License v1.0; no universal hosted token tariff is verified. Free commercial use is revenue-limited around the $10M annual-revenue threshold, with commercial licensing required beyond it. The publisher’s summary and full-text boundary wording differ at exactly $10M; confirm that boundary with Liquid before relying on it. Self-hosting still incurs compute costs.

Best for

On-device / edge
Agents
Cost-sensitive

Watch out

The exact config confirms 131,072 positions, but runtime capacity depends on serving configuration. Native tool calling is documented. Attribution and licence notices must be retained on redistribution; qualified nonprofit noncommercial/research use has a specific exemption. The public benchmark table estimates an index of 8 (asterisk), so it is withheld from the independently matched score display. Throughput figures use H100 SXM5, SGLang 0.5.16, BF16, 1,024 input/up to 256 output tokens, and three-run averages.

Source receipts

Catalog figures for this model were checked against the following sources.

Liquid model release and extension; original metadata dates August 4 (accessed 2026-10-25)
Liquid exact model documentation (accessed 2026-10-25)
Liquid checkpoint architecture and usage (accessed 2026-10-25)
Liquid raw 131,072-position configuration (accessed 2026-10-25)
LFM Open License v1.0: full terms, threshold and nonprofit exception (accessed 2026-10-25)
Liquid current licensing pricing; no universal inference tariff (accessed 2026-10-25)
Artificial Analysis retained estimated row; not an independent measured score (accessed 2026-10-01)

Common questions

Liquid LFM2.5-2.6B

Answered from the verified figures on this page rather than general guidance.

What is Liquid LFM2.5-2.6B's context window?

Liquid LFM2.5-2.6B accepts about 131K tokens of context. That only matters if you routinely send very long documents, large codebases, or multi-turn histories that approach that limit.

What is Liquid LFM2.5-2.6B best for?

Liquid LFM2.5-2.6B is a budget tier from Liquid AI. It suits on-device / edge, agents, cost-sensitive. The exact config confirms 131,072 positions, but runtime capacity depends on serving configuration. Native tool calling is documented. Attribution and licence notices must be retained on redistribution; qualified nonprofit noncommercial/research use has a specific exemption. The public benchmark table estimates an index of 8 (asterisk), so it is withheld from the independently matched score display. Throughput figures use H100 SXM5, SGLang 0.5.16, BF16, 1,024 input/up to 256 output tokens, and three-run averages.

Can I self-host Liquid LFM2.5-2.6B?

Liquid LFM2.5-2.6B publishes open weights, but self-hosting depends on the licence, hardware footprint, quantisation quality, and serving stack. A hosted API is often cheaper until you have measured throughput and concurrency on your own hardware.

Model comparison catalog Benchmark library