Dedicated model · Available as managed deployment
Meta's small Llama 3.2 models (1B and 3B instruction-tuned) for fast, cheap text tasks that do not need a large model: classification, extraction, summaries, on-device style workloads on a server. Available as a managed deployment: validated on AxForge hardware and deployed on a DGX Spark for your traffic only, as an OpenAI-compatible endpoint on hardware only you use, hosted in the EU.
Why Llama 3.2
| Small, fast, cheap | 1B and 3B parameters: thousands of requests per hour on a single dedicated machine, with latency a large model cannot match. |
|---|---|
| The right size for the job | Routing, tagging, extraction, short summaries, guardrails: the tasks where a small model at high volume beats a big one at any price. |
| Same endpoint as the big ones | OpenAI-compatible on your own hardware, so the small model slots into a pipeline next to Llama 3.1 or Qwen without a different SDK. |
Specifications
| Model | Llama 3.2 by meta-llama |
|---|---|
| Modalities | Text |
| Sizes | 1.2B, 3.2B |
| Released | 2024-09 |
| Popularity | 6.7M downloads / month on HuggingFace |
| Licence | Open, with conditions (llama3.2); AxForge deploys under it and tells you what applies |
| Availability | Available as managed deployment, operated by AxForge on hardware reserved for you |
| Hardware | NVIDIA DGX Spark (GB10, 128 GB unified memory), owned and operated by AxForge |
| Region | Málaga, Spain (eu-es-1) |
| Pricing | Hardware from €0.85/hour, rental term: Hour, week, month or year. Managed service quoted per deployment. Request deployment and we confirm both in writing. |
We publish only numbers we measure ourselves, and we haven't benchmarked this model on our nodes yet. For quality benchmarks, see the model card in the catalogue and the official card on HuggingFace.
Deployment
| System | A dedicated NVIDIA DGX Spark (GB10, 128 GB unified memory), reserved for you |
|---|---|
| Commercial model | Hour, week, month or year. Hardware from €0.85/hour, managed service quoted per deployment |
| Endpoint | OpenAI-compatible /v1 on your own machine, only your traffic |
| Location | Hosted in the EU: Málaga, Spain (eu-es-1) |
| Data handling | Zero prompt retention, same policy as the serverless API |
The model is validated on AxForge hardware and deployed on your system. After deployment, your endpoint speaks the OpenAI API with the model name you receive with the deployment.
Data & privacy
Prompts and completions are processed in memory in the EU, not written to disk, not logged, not retained, and never used to train anything. We keep only request metadata (token counts, timestamps, status) for billing and operations. The full policy is at axforge.ai/privacy.
FAQ
Not on the serverless API. It is available as a managed deployment: validated on AxForge hardware and deployed on a DGX Spark for your traffic only. The serverless API serves Qwen3.8 27B.
High-volume, low-latency text tasks: classification, extraction, routing, short summaries and guardrails. For general chat quality look at Llama 3.1 or Qwen3.
Yes, a dedicated DGX Spark has room for a small model next to a large one; AxForge configures both behind your endpoint.
We publish only numbers we measure ourselves, and we haven't benchmarked this model on our nodes yet. For quality benchmarks, see the official model card.
A dedicated NVIDIA DGX Spark (GB10, 128 GB unified memory) rented by the hour, week, month or year, running Llama 3.2 behind an OpenAI-compatible endpoint on your own machine, hosted in the EU with zero prompt retention.
Hardware from €0.85/hour, by the hour, week, month or year; the managed service is quoted per deployment. Request deployment and we confirm both in writing.
Explore