🔗 Day-0 Support for LFM2.5-1.2B‑Thinking
First reasoning model from LiquidAI! More details: https://huggingface.co/LiquidAI/LFM2.5-1.2B-Thinking
LFM2.5-1.2B-Thinking delivers strong improvements in math reasoning, instruction following, and tool use, matching or exceeding Qwen3-1.7B on most reasoning benchmarks despite using ~40% fewer parameters
Model tag in FLM: lfm2.5-tk:1.2b
Run it in CLI Mode with:
flm run lfm2.5-tk:1.2bRun it in Server Mode with:
flm serve lfm2.5-tk:1.2b📊 Performance at a Glance
Kraken (Ryzen AI 340 / 350)
| Device | Inference | Framework | Model | 4K-Token Prefill Speed (tok/s) | Peak Decoding Speed (tok/s) | Memory (Full Context) |
|---|---|---|---|---|---|---|
| AMD Ryzen AI 7 HX350 | NPU | FastFlowLM | LFM2.5-1.2B-Thinking | 2032 | 63+ | 1.6 GB (full context) |
| AMD Ryzen AI 5 HX340 | NPU | FastFlowLM | LFM2.5-1.2B-Thinking | 2032 | 63+ | 1.6 GB (full context) |
Decoding Speed vs. context length:
- 59 tok/s @ 4K context
- 52 tok/s @ 16K context
Strix/Strix Halo (Ryzen AI 360 and above)
| Device | Inference | Framework | Model | 4K-Token Prefill Speed (tok/s) | Peak Decoding Speed (tok/s) | Memory (Full Context) |
|---|---|---|---|---|---|---|
| AMD Ryzen AI 395+ | NPU | FastFlowLM | LFM2.5-1.2B-Thinking | 2226 | 60+ | 1.6 GB (full context) |
| AMD Ryzen AI 9 HX370 | NPU | FastFlowLM | LFM2.5-1.2B-Thinking | 2226 | 60+ | 1.6 GB (full context) |
Decoding Speed vs. context length:
- 54 tok/s @ 4K context
- 49 tok/s @ 16K context
Detailed benchmarks: https://fastflowlm.com/docs/benchmarks/lfm2_results/
Demo: 💧LFM2.5-1.2B-Thinking (LiquidAI) — 100% Powered by AMD Ryzen™ AI NPU
🆕 Model Tag Change
To avoid confusion between variants:
| Old Tag | New Tag |
|---|---|
lfm2.5:1.2
| lfm2.5-it:1.2 (original, instruct)
|