Brief-details: StableLM Zephyr 3B is a 2.8B parameter instruction-tuned LLM optimized for chat and general text generation, achieving strong performance on MT-Bench (6.64) and various benchmarks.
Brief-details: An anime-focused text-to-image diffusion model specializing in moe character generation, with extensive version history and style improvements across multiple iterations.
Brief Details: A massive 176B parameter multilingual language model trained on 46 languages, using 384 A100 GPUs. Notable for open collaboration and environmental consciousness.
Brief Details: CodeGeeX2-6B is a powerful multilingual code generation model with 6B parameters, built on ChatGLM2 architecture, supporting 100+ programming languages and achieving state-of-the-art performance.
BRIEF DETAILS: Advanced 34B parameter chat model from 01-ai, trained on 3.6T tokens with strong performance in coding, math, and reasoning. Features 4K-32K context lengths.
Brief Details: Extended context Llama-3 model (262k tokens) with 8B parameters, optimized for instruction-following and long-context processing. Meta's base with Gradient AI improvements.
Brief-details: A powerful 34B parameter code generation model fine-tuned on Python tasks, achieving 69.5% pass@1 on HumanEval, surpassing GPT-4's performance of 67%. Built for high-quality code completion and generation.
Brief-details: A powerful 34B parameter LLM fine-tuned on Yi-34B with 200K context length, specialized in multi-turn conversations and complex summaries. Features Amplify-instruct methodology.
Brief-details: Multilingual GPT model with 1.3B parameters supporting 61 languages, trained on Wikipedia and MC4 corpus using Deepspeed and Megatron frameworks.
Brief-details: OpenHermes-2-Mistral-7B is a state-of-the-art Mistral-7B fine-tune trained on 900k GPT-4 entries, using ChatML format and achieving superior benchmark performance.
Brief Details: A powerful 70B parameter LLM fine-tuned from LLaMA-2, achieving top rankings on HuggingFace's Open LLM leaderboard with strong performance in reasoning and instruction-following tasks.
Brief-details: A 7B parameter uncensored language model based on Mistral, trained on Dolphin dataset for enhanced compliance and creativity. Apache 2.0 licensed.
Brief-details: FastSpeech2 text-to-speech model trained on LJSpeech dataset, offering single-speaker female voice synthesis in English. Built by Facebook using Fairseq framework.
Brief-details: Triplex is a 3.82B parameter LLM specialized in knowledge graph construction, offering 98% cost reduction vs GPT-4 while maintaining high accuracy.
BRIEF DETAILS: 70B parameter GPTQ-quantized Llama2 chat model optimized for dialogue. Features multiple quantization options and extensive GPU compatibility.
Brief-details: A powerful 3B parameter language model with 8k context length, matching 7B model performance. Features ALiBi position embeddings and SwiGLU activation, trained on SlimPajama-627B dataset.
Brief Details: Zephyr 141B - Advanced MoE language model with 141B params (39B active). Fine-tuned using ORPO on Mixtral-8x22B. Strong MT-Bench (8.17) and IFEval (65.06) performance.
Brief-details: Yi-VL-34B is a state-of-the-art bilingual vision-language model with 34B parameters, supporting multi-round image-text conversations and achieving top performance in MMMU benchmarks.
Brief-details: A comprehensive guide for Stable Diffusion implementation, covering model setup, prompting, ControlNet, and Lora training with practical examples and tips
Brief Details: DePlot is a 282M parameter visual reasoning model that translates plots/charts to text for LLM processing, supporting 5 languages with state-of-the-art performance in chart QA.
Brief Details: A powerful 13B parameter Russian language model trained on 300GB+ diverse data, achieving 8.8 perplexity. Supports both Russian and English text generation.
