Tele-FLM

An open-source multilingual large language model with 52 billion parameters

PremiumNewProductProgrammingLarge Language ModelMultilingual Support

Tele-FLM (also known as FLM-2) is a 52-billion parameter open-source multilingual large language model with a stable and efficient pre-training paradigm and enhanced fact-checking capabilities. Based on a decoder-only transformer architecture, it has been trained on approximately 2 trillion tokens. Tele-FLM exhibits superior performance compared to models of similar size, sometimes even surpassing larger ones. Besides sharing the model weights, we also provide core design, engineering practices, and training details, hoping they will benefit both the academic and industrial communities.

Decoder-only transformer architecture-based model
optimized for fact-checking capabilities.
Supports multiple languages
including English and Chinese.
Provides core design and engineering practices for easy community use and learning.
Training data covers multiple domains
encompassing a wide range of knowledge.
Utilizes 3D parallel training techniques to enhance training efficiency.
Demonstrates good performance on multiple benchmark datasets.

Tele-FLM is primarily aimed at developers and researchers who need to process and generate multilingual text
especially professionals in the field of natural language processing seeking efficient and high-performing models.

Used for generating concise summaries of text in specific domains.
Provides accurate information retrieval and answering capabilities in question-answering systems.
Serves as a backend for chatbots
delivering a smooth conversational experience.

1. Import the torch and transformers libraries.
2. Load the tokenizer and model from the pre-trained model using AutoTokenizer and AutoModelForCausalLM.
3. Convert the input text into a format understandable by the model using the tokenizer.
4. Send the converted input data to the model's device.
5. Generate text using the model.generate method.

Visit

Tele-FLM Visit Over Time

Monthly Visits

20899836

Bounce Rate

46.04%

Page per Visit

5.2

Visit Duration

00:04:57

Tele-FLM Visit Trend

Tele-FLM Visit Geography

Tele-FLM Traffic Sources

Tele-FLM Alternatives

Mistral-Nemo-Instruct-2407 — Large language model, supports multilingual and code data

Programming•Large language model•Multilingual support

444

OpenBioLLM-Llama3-8B — An open-source large language model specifically designed for the biomedical field

OpenSource•Biomedical•Natural Language Processing

378

EXAONE-3.5-32B-Instruct-GGUF — A multilingual, high-performance large language model developed by LG AI Research.

Productivity•Large language model•Multilingual support

CosyVoice Speech Generation Model 2.0-0.5B — Efficient, multilingual speech synthesis model

Music•Speech Synthesis•Artificial Intelligence

180

InternVL 2.5 — Open-source multimodal large language model series

Productivity•multimodal•large language model

OLMo 2 7B — A large language model with 7 billion parameters, enhancing natural language processing capabilities.

Programming•Large Language Model•Natural Language Processing

Skywork-o1-Open-PRM-Qwen-2.5-7B — High-performance AI model that enhances reasoning task capabilities.

Programming•AI Reasoning•Machine Learning

BEXI.ai — Free natural language AI personalization tool

Writing•AI text humanization•Natural language processing

Mistral-Large-Instruct-2411 — A large language model with 123 billion parameters, equipped with advanced reasoning and coding capabilities.

Programming•Large Language Model•Reasoning

Pixtral-Large-Instruct-2411 — A 124B-parameter multimodal large language model.

Productivity•Multimodal•Large Language Model

102

Ultravox.ai — Next-generation voice AI, creating AI voice agents for natural communication.

Programming•AI Voice•Natural Language Processing

222

Ministral-8B-Instruct-2410 — A high-performance language model that supports local intelligence and on-device computation.

Programming•Large Language Model•Local Intelligence

Gaia-ml — No coding required; quickly build neural machine translators.

Productivity•Machine Translation•Artificial Intelligence

Llama-3.2-3B — Multilingual Large Language Model

Productivity•Artificial Intelligence•Machine Learning

222

Qwen2.5 — An open-source large language model that supports multilingual and specialized domain applications.

Productivity•Artificial Intelligence•Natural Language Processing

240

Reflection Llama-3.1 70B — The world's leading open-source large language model

Productivity•Large Language Model•Natural Language Processing

294

OLMoE-1B-7B — An efficient open-source large language model.

Productivity•Large Language Model•Open Source

C4AI CommandR 08-2024 — A high-performance generation model with 3.5 billion parameters

Productivity•Large Language Model•Multilingual Support

C4AI Command R+ 08-2024 — A multilingual advanced dialogue generation model with 104 billion parameters.

Productivity•Artificial Intelligence•Machine Learning

Meta-Llama-3.1-405B-Instruct — A multilingual large language model optimized for conversational contexts.

Programming•Large Language Model•Conversational System

288

Meta-Llama-3.1-70B-Instruct — A large multilingual dialogue generation model with 7 billion parameters.

Programming•Large Language Model•Dialogue Generation

348

Llama3 — Large language model supporting various parameter sizes.

Programming•Large language model•Natural language processing

HIX Bypass — The Perfect Disguise for AI-Generated Content

InternationalSelection•AI Detection•Content Rewriting

264

Fish Speech V1.2 — Leading Text-to-Speech Conversion Model

Others•Text-to-Speech•Multilingual Support

990

Index-1.9B-Pure — A lightweight large language model focused on text generation.

Programming•Text Generation•Natural Language Processing

126

Index-1.9B — Index-1.9B is a lightweight, open-source large language model developed independently by Bilibili.