# femto-jev decision model by igorktech

> femto-jev is a small text and speech decision model with open weights. Check its separate speech frontend, reported tests and out-of-list limits.

Updated 2026-11-12 (Australia/Adelaide)

## Quick answers

### What is femto-jev by igorktech?

[femto-jev](<https://decisionmodels.ai/models/femto-jev>): A small typed-decision encoder for text, with a separate femto-asr speech frontend.

[femto-jev facts](<https://decisionmodels.ai/models/femto-jev>)

[Source](<https://huggingface.co/igorktech/femto-jev>)

### Who makes femto-jev?

[femto-jev](<https://decisionmodels.ai/models/femto-jev>): igorktech

[femto-jev facts](<https://decisionmodels.ai/models/femto-jev>)

[Source](<https://huggingface.co/igorktech/femto-jev>)

### Are femto-jev's weights open?

[femto-jev](<https://decisionmodels.ai/models/femto-jev>): Public weights and single-file Python inference; no text generation Licence: Apache-2.0, with upstream notices

[femto-jev facts](<https://decisionmodels.ai/models/femto-jev>)

[Source](<https://huggingface.co/igorktech/femto-jev>)

By igorktech · Hugging Face weights, 15:42:55 UTC: 2026-11-11 · OPEN

A small typed-decision encoder for text, with a separate femto-asr speech frontend. It scores options in a forward pass per question. The maker’s speech tests show that audio frames help on familiar recording conditions but add no benefit on unseen phone speech.

## Facts

- Size: Maker reports 7.7M decision-model parameters and 30 MB fp32 weights; speech frontend adds 2.39M parameters
- Inputs: The maker reports text or 16 kHz mono speech input through femto-asr; choice, yes/no and score questions
- Access: Public weights and single-file Python inference; no text generation
- Licence: Apache-2.0, with upstream notices
- Limits: English; out-of-list spoken commands are not reliably rejected. Confidence is not an out-of-distribution rejection guarantee

## Claims and sources

- **VENDOR-REPORTED**: The maker reports 73.2% accuracy across 400 typed-decisions test cases containing 2,000 decisions. The model trained on that dataset’s training split. SLURP accuracy is 71.4% with transcript plus frames, against 58.6% with transcript alone; on unseen MInDS-14 phone speech the figures are 55.8% and 56.1%. [source](<https://huggingface.co/igorktech/femto-jev>)

## Links

- [Source: Card and weights](<https://huggingface.co/igorktech/femto-jev>)
- [Source: Inference code](<https://huggingface.co/igorktech/femto-jev/blob/main/femto_jev.py>)
- [Source: Licence](<https://huggingface.co/igorktech/femto-jev/blob/main/LICENSE>)
- [Source: Configuration](<https://huggingface.co/igorktech/femto-jev/blob/main/rl_agent_config.json>)

## Related projects

No related projects are listed in the registry.