LFM2.5 8B A1B
LFM2.5 8B A1B is a released rag, agents, and long context model with open-weight and 128K context; evaluate it while provider pricing coverage matures.
- Family
- LFM-2.5
- Released
- 2026-05-28
- Context
- 128K
- Max output
- 8K
- Parameters
- 8.3B
- Architecture
- Mixture of Experts
- Specialization
- general
- Openness
- Open weights
- License
- LFM Open License v1.0Commercial use: conditional
- Weights
- Available
- Code
- Unknown
- Training
- Pretrained
No tracked provider token pricing is available yet.
About
LFM2.5-8B-A1B is Liquid AI's latest on-device mixture-of-experts model, succeeding LFM2-8B-A1B. It has 8.3B total parameters with approximately 1.5B active per token (the A1B label uses a rounded ~1B figure). The architecture combines 18 double-gated LIV convolutional layers with 6 GQA attention layers, trained on 38 trillion tokens. The context window expands to 128K tokens (up from 32K in the predecessor). It is a reasoning model that generates explicit chain-of-thought steps before producing its final answer, making reasoning tokens cheap due to the MoE design. Strong tool-calling, function-calling, and instruction-following capabilities make it well-suited for agentic workflows on edge hardware.
Provider price ladder
No tracked provider token pricing is available for this model yet.
Capabilities
Benchmark peer barsfor Agents
Benchmark scores(6)
| Benchmark | Score | Version | Evaluation | Source |
|---|---|---|---|---|
| Instruction-Following Evaluation | 91.8 | —Observed 2026-05-28 | — | Source |
| MATH-500 | 88.8 | MATH-500 (accuracy)Observed 2026-06-07 | — | Source |
| AIME 2025 | 42.5 | —Observed 2026-05-28 | — | Source |
| Google-Proof Q&A | 34.4 | GPQA Diamond (accuracy)Observed 2026-06-07 | — | Source |
| MMLU PRO | 50.5 | Third-party evaluation (accuracy)Observed 2026-06-07 | — | Source |
| τ-bench | 88.1 | τ² Telecom benchmark (accuracy)Observed 2026-06-07 | — | Source |
Migration checks
No linked migration route is available for this model yet.
No tracked provider token pricing is available yet.