What are the hardware requirements for Longcat 10.7B?

Longcat 10.7B requires approximately 21.5 GB of VRAM and supports a context window of 32K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Longcat 10.7B and how large is it?

Longcat 10.7B is developed by SanjiWatsuki, a model with 10.7b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Longcat 10.7B perform on standard benchmarks?

Longcat 10.7B has the following published scores: MMLU 61.77. Compare against reference models on this page or on the LLM Explorer leaderboards.

Longcat 10.7B by SanjiWatsuki — VRAM 21.5GB, 32K context

Name: Longcat 10.7B
Author: SanjiWatsuki

Longcat 10.7B is an open-source language model by SanjiWatsuki. Features: 10.7b LLM, VRAM: 21.5GB, Context: 32K, License: apache-2.0, LLM Explorer Score: 0.11, Arc: 64.6, HellaSwag: 85.9, MMLU: 61.8, GSM8K: 46.1.

Base model:finetune:rishiraj/c... Base model:rishiraj/catppt-bas... En Endpoints compatible Merge Mistral Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/SanjiWatsuki/longcat-10.7B

Longcat 10.7B Parameters and Internals

Model Type

merge, Interwoven Depth Up-Scaling, Mistral-7B

Additional Notes

Do not attempt to merge with other 10.7Bs due to layer mismatch. Developed with humor referring to creating a 'longcat'.

Supported Languages

en (full)

Training Details

Methodology:

Merging using a new process called Interwoven Depth Up-Scaling

Model Architecture:

Mistral architecture with Interwoven Depth Up-Scaling

LLM Name	Longcat 10.7B
Repository 🤗	https://huggingface.co/SanjiWatsuki/longcat-10.7B
Base Model(s)	CatPPT Base rishiraj/CatPPT-base
Model Size	10.7b
Required VRAM	21.5 GB
Updated	2026-07-30
Maintainer	SanjiWatsuki
Model Type	mistral
Model Files	10.0 GB: 1-of-3 10.0 GB: 2-of-3 1.5 GB: 3-of-3
Supported Languages	en
Model Architecture	MistralForCausalLM
License	apache-2.0
Context Length	32768
Model Max Length	32768
Transformers Version	4.36.2
Tokenizer Class	LlamaTokenizer
Vocabulary Size	32000
Torch Data Type	bfloat16

Rank the Longcat 10.7B 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 Longcat 10.7B

Best Alternatives	Context / RAM	Downloads	Likes
Barely Regal 10.7B	32K / 21.5 GB	19	0
Quintellect 10.7B	32K / 21.4 GB	22	3
Chikuma 10.7B	32K / 21.4 GB	131	5
PiVoT 10.7B Mistral V0.2	32K / 21.4 GB	76	5
Multi Verse Model 10.7B	32K / 21.5 GB	2	1
PrimaSumika 10.7B 128K	32K / 21.5 GB	3	1
... Loyal Mistral Maid 32K V0.2 A	32K / 21.5 GB	3	1
Chikuma 10.7B V2	32K / 21.4 GB	10	0
PiVoT 10.7B Mistral V0.2 RP	32K / 21.4 GB	67	7
Mark1 Revision 10.7B	32K / 42.9 GB	16	1

Note: green Score (e.g. "73.2") means that the model is better than SanjiWatsuki/longcat-10.7B.