Model library / superwhisper
S1-mini is a 0.6 B-parameter text normaliser that cleans raw English ASR output into properly punctuated, capitalised written text.
Parameters in total
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
Superwhisper's S1-mini is a 0.6 B-parameter text normaliser that rewrites raw speech-to-text transcripts into clean written English, removing fillers, correcting false starts, and applying punctuation, capitalization and appropriate rendering of numbers, dates, times, currency and email addresses as described by the publisher for downstream dictation applications.
The model contains 752 million unique parameters and is stored in bfloat16 format. Its context window spans 40,960 tokens. The quantised GGUF build is 462 MiB and runs comfortably on a laptop CPU; GPU use is possible via `device_map="auto"`. The licence is listed as "other", and the model type is qwen3 without expert layers.
Specification
| Repository | superwhisper/s1-mini |
| Publisher | superwhisper |
| Published | 12 August 2026 |
| Architecture | Qwen3ForCausalLM |
| Model type | qwen3 |
| Parameters | 751,632,384 |
| Layers | 28 |
| Hidden size | 1,024 |
| Attention heads | 16, 8 key/value heads |
| Context length | 40,960 tokens |
| Vocabulary | 151,936 tokens |
| Weight format | bfloat16 |
| Licence | other |
| Base model | Qwen/Qwen3-0.6B |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 1.4 GiB | Full precision as released |
| fp8 | 0.7 GiB | 8-bit, near-lossless on most models |
| int4 | 0.4 GiB | 4-bit, smallest footprint |
The cache costs 112 KiB per token at 16-bit, so the full 40,960-token context of one request needs about 4.4 GiB. Concurrency multiplies that number, not the weights.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | 1 | 17.7 GiB | the full 40,960 tokens |
| RTX 5090 32GB | 1 | 24.9 GiB | the full 40,960 tokens |
| L40S 48GB | 1 | 39.3 GiB | the full 40,960 tokens |
| A100 80GB | 1 | 68.1 GiB | the full 40,960 tokens |
| H100 80GB | 1 | 68.1 GiB | the full 40,960 tokens |
| RTX PRO 6000 96GB | 1 | 82.5 GiB | the full 40,960 tokens |
| H200 141GB | 1 | 123.0 GiB | the full 40,960 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve superwhisper/s1-mini \ --max-model-len 40960 \ --gpu-memory-utilization 0.90
Questions
The weights take about 1 GiB at the released precision. That fits on one RTX 4090 24GB, which leaves roughly 18 GiB for the key-value cache.
The repository declares other. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 40,960 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.
Also in the library
K2-Horizon-0.9B is a 1.1 billion-parameter dense decoder-only model with a 128 k token context window.
Nanbeige4.2-3B is a compact agentic language model designed for reasoning, tool use and personal-assistant tasks.
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.