Aplomb 1
EmpirioLabs' first decision model, a 5.3B fine-tune of Qwen3.5-4B plus an audio encoder. Reads text, JSON, images, video and audio and answers up to 128 yes/no, choice, score and tool-selection questions per request, with a probability per option. Gated weights; hosted API.
The weights are click-through gated. The EmpirioLabs Model License allows research, evaluation and internal commercial use by organisations under US$1M revenue; offering it as a hosted service, at any size, needs a written commercial agreement, and distilling from its outputs is banned. So it is listed here only, not mirrored, packaged or served by System One Models. The hosted Decisions API (generally available, $0.02 per million input tokens, output free) also accepts OpenAI, Anthropic and Gemini request formats. The audio encoder comes from Qwen3-Omni-30B-A3B-Instruct. On the maker's own average over seven sets run with the Intern-Decision evaluation bundle it ties Jev (88.7%) and trails Intern-Decision-4B (89.9%) in the maker's table. Calibration is claimed only from a 100-item pilot, with no ECE; audio results are uneven (ESC-50 8.4%); the card does not say whether each question is one forward pass. EmpirioLabs' own run of the third-party Decision Index 0.2.1 kit gives 44.86, which is not a board placement.
What it decides
- choice — picks one option from a set
- score — places the input on an ordered scale
- noul — answers a yes/no question with one probability
- classify — assigns a category from a fixed taxonomy
- route — sends the input to one of several destinations
At a glance
| Parameters | 5.3B |
| Base model | Qwen/Qwen3.5-4B |
| Maker | EmpirioLabs |
| Released | 2026-09-29 |
| License | EmpirioLabs Model License 1.0 (free under US$1M revenue; hosting needs a licence; non-commercial for us) |
| Reported accuracy | 88.7% |
| Reported latency | about 15 ms model time per short question, 35 ms with an image, about 3 s for a 1M-token state (hardware not stated) |
Hosted API
Served by EmpirioLabs — $0.02/MTok input, $0/MTok output. Get access · API docs.
Get the weights
pip install systemonemodels
systemone pull empiriolabs/aplomb-1
The files are served from the maker's Hugging Face repository, empiriolabsai/aplomb-1, and verified against the checksums recorded here.
Read more
This page was opened by System One for EmpirioLabs, who can claim the organisation and take it over at any time.