LFM2 · Liquid AI
Fine-tuning LFM2 8B-A1B
What does it take to fine-tune LFM2 8B-A1B?
LFM2 8B-A1B needs 18 GB for half-precision LoRA training and 12 GB in four-bit, and 16 GB to serve. It supports supervised fine-tuning, under the LFM Open licence. The cheapest qualifying class is RTX 3080 at $0.09 per GPU-hour.
Eight billion parameters with only 1.5 billion active per token — the most aggressive sparsity ratio in the catalogue. The practical effect is 8B-class knowledge at closer to 1.5B-class inference cost, provided you can afford to keep all eight billion resident.
What to know before choosing it
Eight billion parameters with 1.5 billion active is the most aggressive sparsity ratio in the catalogue. In practice that means 8B-class knowledge at closer to 1.5B-class inference cost — provided you can afford to keep all eight billion resident, which is the part people forget.
It is therefore a serving-economics play rather than a training-economics one. On a busy endpoint the throughput advantage is real; on an intermittent one you are paying 8B memory for 1.5B of work.
Specification
| Parameters | 8.3B1.5B active per token |
|---|---|
| Architecture | Mixture of experts |
| Size tier | Mid |
| LoRA training memory | 18 GBHalf precision, frozen base |
| QLoRA training memory | 12 GBFour-bit base, higher-precision adapter |
| Serving memory | 16 GBHalf precision, before attention cache |
| Objectives | SFT |
| Licence | LFM Open |
| Repository | LiquidAI/LFM2-8B-A1B |
What it costs to train
Every class with enough memory for four-bit training, cheapest first. Total cost is the rate multiplied by wall time, so the cheapest rate is not always the cheapest run — a faster class that finishes sooner frequently wins.
| GPU class | VRAM | Training / hr | Serving / hr | Fits |
|---|---|---|---|---|
| RTX 3080 | 12 GB | $0.09 | $0.11 | QLoRA only |
| RTX 4000 Ada | 20 GB | $0.09 | $0.11 | LoRA and QLoRA |
| L4 | 24 GB | $0.17 | $0.20 | LoRA and QLoRA |
| RTX 3090 | 24 GB | $0.22 | $0.26 | LoRA and QLoRA |
| RTX 4090 | 24 GB | $0.38 | $0.43 | LoRA and QLoRA |
| A40 | 48 GB | $0.42 | $0.48 | LoRA and QLoRA |
| RTX 6000 Ada | 48 GB | $0.61 | $0.71 | LoRA and QLoRA |
| A6000 | 48 GB | $0.65 | $0.75 | LoRA and QLoRA |
| L40S | 48 GB | $0.78 | $0.90 | LoRA and QLoRA |
| A100 40 GB | 40 GB | $1.17 | $1.35 | LoRA and QLoRA |
| A100 80 GB | 80 GB | $1.68 | $1.94 | LoRA and QLoRA |
| H100 80 GB | 80 GB | $2.59 | $2.99 | LoRA and QLoRA |
| H200 | 141 GB | $4.55 | $5.25 | LoRA and QLoRA |
Half-precision LoRA needs 18 GB, so the cheapest class for it is RTX 4000 Ada at $0.09 per hour. Below that, training has to be quantised.
Serving fits on RTX 4000 Ada at $0.11 per hour — before the attention cache, which grows with context length and concurrency.
Good starting point for
- Maximum throughput per dollar of serving
- Knowledge-heavy tasks with tight latency budgets
This is a mixture-of-experts model
8.3B total parameters with 1.5B active per token. Memory scales with the total; inference compute scales with the active count. That makes it attractive when memory is cheaper than compute for your workload, and unattractive when the reverse holds. Quoting only one of the two numbers is how these models get misrepresented in both directions.
Other sizes in this family
| Model | Params | QLoRA | Objectives |
|---|---|---|---|
| LFM2 350M | 350M | 2 GB | SFT |
| LFM2 700M | 700M | 3 GB | SFT |
| LFM2 1.2B | 1.2B | 4 GB | SFT |
| LFM2 2.6B | 2.6B | 6 GB | SFT |
| LFM2 24B-A2B | 24B | 28 GB | SFT |
Comparable sizes elsewhere
Frequently asked questions
How much VRAM does it take to fine-tune LFM2 8B-A1B?
18 GB for half-precision LoRA and 12 GB for four-bit QLoRA. Serving needs 16 GB. Preference tuning roughly doubles the training figure, because a frozen reference model is held alongside the one being trained.
What is the cheapest way to fine-tune LFM2 8B-A1B?
Four-bit QLoRA on RTX 3080 at $0.09 per GPU-hour is the cheapest class that meets the 12 GB threshold. Note that quantised training is slower per step, so a faster class sometimes costs less over the whole run.
Can I download the weights after fine-tuning LFM2 8B-A1B?
Yes. Every finished run exposes its trained weights for download, and publishing to a model hub is a single call with a generated model card recording the base model and version the adapter applies to.
What licence does LFM2 8B-A1B carry?
LFM Open. The licence follows the fine-tune — a derivative inherits the base model’s terms, and those terms pass to anyone you give the model to.
Last verified 6 August 2026. Memory thresholds are the platform's own admission limits.
Fine-tune LFM2 8B-A1B
Upload a dataset, forecast the run, and see the cost before any compute is leased.