Model comparison
Conv-TasNet Libri1Mix (vokra) vs GPT-5
Capability radar, value bars, and side-by-side posture.
Conv-TasNet Libri1Mix (vokra)
Not scored
value score
GPT-5
94
value score
Conv-TasNet Libri1Mix (vokra) ctx
—
GPT-5 ctx
400K
Capability radar
Value · context · multimodal · openness · speed posture
Value head-to-head
- GPT-594
Attribute tape
| Field | Conv-TasNet Libri1Mix (vokra) | GPT-5 |
|---|---|---|
| Developer | Vokra | OpenAI |
| Context | — | 400K |
| Modalities | Text, Code, Image, Audio | |
| Openness | — | Proprietary Model |
| Speed | — | Fast |
| Price | — | Mid-High |
| Value score | — | 94 |
| Summary | A Conv-TasNet speech enhancement model trained on the Libri1Mix enh_single task at 16 kHz, capable of separating and cleaning speech from noisy single-channel audio mixtures entirely in the time domain. It is derived from the Asteroid librimix recipe and closely mirrors the JorisCos/ConvTasNet_Libri1Mix_enhsingle_16k reference artifact. | OpenAI frontier multimodal model. |