What are the hardware requirements for TwIL 8B?

TwIL 8B requires approximately 5.3 GB of VRAM and supports a context window of 128K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed TwIL 8B and how large is it?

TwIL 8B is developed by webAI-Official, a model with 8b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does TwIL 8B perform on standard benchmarks?

TwIL 8B has the following published scores: LMSYS ELO 1290. Compare against reference models on this page or on the LLM Explorer leaderboards.

TwIL 8B by webAI-Official — VRAM 5.3GB, 128K context

Name: TwIL 8B
Author: webAI-Official

TwIL 8B is an open-source language model by webAI-Official. Features: 8b LLM, VRAM: 5.3GB, Context: 128K, License: other, Quantized, Instruction-Based, ELO: 1290.

Arxiv:2607.02770 Base model:ibm-granite/granite... Base model:quantized:ibm-grani... Code Conversational En Endpoints compatible Fp16 Gguf Granite Granite-4.2 Grpo Instruct Instruction-following Math Q4 Q8 0 Quantized Reasoning Region:us Safetensors Twil

Model Card on HF 🤗: https://huggingface.co/webAI-Official/TwIL-8B

TwIL 8B Benchmarks

LMSys ELO 1290 vs 1272 (so35) 1.4%

^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)

TwIL 8B Parameters and Internals

LLM Name	TwIL 8B
Repository 🤗	https://huggingface.co/webAI-Official/TwIL-8B
Base Model(s)	ibm-granite/granite-4.2-8b ibm-granite/granite-4.2-8b
Model Size	8b
Required VRAM	5.3 GB
Updated	2026-11-05
Maintainer	webAI-Official
Model Type	granite
Instruction-Based	Yes
Model Files	17.6 GB 5.3 GB 5.3 GB 9.3 GB 17.6 GB
Supported Languages	en
GGUF Quantization	Yes
Quantization Type	fp16\|gguf\|q4\|q8_0\|q4_k
Model Architecture	GraniteForCausalLM
License	other
Context Length	131072
Model Max Length	131072
Transformers Version	5.5.3
Tokenizer Class	GPT2Tokenizer
Padding Token	<\|im_end\|>
Vocabulary Size	100352
Errors	replace

Rank the TwIL 8B 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 TwIL 8B

Best Alternatives	Context / RAM	Downloads	Likes
Granite 3.3 8B Instruct GGUF	128K / 1.9 GB	521	7
Granite 3.3 8B Instruct Fp16	128K / 16.4 GB	63	3
Granite 3.3 8B Instruct 8bit	128K / 8.6 GB	53	2
...anite 3.2 8B Instruct Bnb 4bit	128K / 4.5 GB	186	3
Granite 3.1 8B Instruct	128K / 16.4 GB	138888	168
Granite 3.1 8B Reasoning	128K / 16.4 GB	5	1
Granite 3.1 8B Instruct	128K / 16.4 GB	185	1
Granite 3.2 8B Instruct	128K / 16.4 GB	3792	92
Granite 3.3 8B Instruct	128K / 16.4 GB	55103	160
Granite 3.3 8B Instruct FP8	128K / 8.4 GB	6980	3

Note: green Score (e.g. "73.2") means that the model is better than webAI-Official/TwIL-8B.