What are the hardware requirements for S1 Mini?

S1 Mini requires approximately 1.5 GB of VRAM and supports a context window of 40K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed S1 Mini and how large is it?

S1 Mini is developed by superwhisper, a model with 0.6b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does S1 Mini perform on standard benchmarks?

S1 Mini has the following published scores: LMSYS ELO 1337. Compare against reference models on this page or on the LLM Explorer leaderboards.

S1 Mini by superwhisper — VRAM 1.5GB, 40K context

Name: S1 Mini
Author: superwhisper

S1 Mini is an open-source language model by superwhisper. Features: 0.6b LLM, VRAM: 1.5GB, Context: 40K, License: other, LLM Explorer Score: 0.52, ELO: 1337.

Arxiv:2505.09388 Asr Automatic-speech-recognition Base model:finetune:qwen/qwen3... Base model:qwen/qwen3-0.6b Conversational Dictation En Endpoints compatible Inverse-text-normalization Post-processing Punctuation Qwen3 Region:us Safetensors Speech-to-text Text-normalization Truecasing

Model Card on HF 🤗: https://huggingface.co/superwhisper/s1-mini

S1 Mini Benchmarks

LMSys ELO 1337 vs 1272 (so35) 5.1%

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

S1 Mini Parameters and Internals

LLM Name	S1 Mini
Repository 🤗	https://huggingface.co/superwhisper/s1-mini
Base Model(s)	Qwen/Qwen3-0.6B Qwen/Qwen3-0.6B
Model Size	0.6b
Required VRAM	1.5 GB
Updated	2026-08-21
Maintainer	superwhisper
Model Type	qwen3
Model Files	1.5 GB
Supported Languages	en
Model Architecture	Qwen3ForCausalLM
License	other
Context Length	40960
Model Max Length	40960
Transformers Version	4.51.0
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Vocabulary Size	151936
Torch Data Type	bfloat16
Errors	replace

Rank the S1 Mini capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to S1 Mini

Best Alternatives	Context / RAM	Downloads	Likes
Qwen3 0.6B JSON SFT	40K / 1.2 GB	428	1
Qwen3 Reranker 0.6B	40K / 1.2 GB	2508313	377
RIFA Edge 0.6B	40K / 1.2 GB	581	0
Meridian Ac Nano	40K / 1.2 GB	493	1
Qwen3 0.6B JSON SFT	40K / 1.2 GB	88	0
Qwen3 0.6B FP8	40K / 1.1 GB	1581403	62
Qwen3 0.6B JSON SFT	40K / 1.2 GB	1676	0
Blake Haiku 1	40K / 1.2 GB	386	0
Tenorio 0.6B	40K / 1.2 GB	251	2
Quipu 0.6B	40K / 1.2 GB	351	4