Authentication
The following models are available through the Oraicle API:
deepseek-ai/DeepSeek-R1
First-generation reasoning model trained via large-scale reinforcement learning (RL), achieving performance comparable to OpenAI-o1 across math, code, and reasoning tasks
🧠 Most popular choice for advanced reasoning and problem-solving tasks
deepseek-ai/DeepSeek-R1-Distill-Llama-70B
Fine-tuned based on open-source Llama 3.3 70B using samples generated by DeepSeek-R1, with slightly modified configs and tokenizers
⚖️ Excellent balance of performance and cost for enterprise applications
deepseek-ai/DeepSeek-R1-Distill-Qwen-32B
Fine-tuned based on open-source Qwen-32B using samples generated by DeepSeek-R1, with slightly modified configs and tokenizers
💼 Popular for commercial deployments seeking cost efficiency
Qwen/QwQ-32B
Medium-sized reasoning model capable of achieving competitive performance against state-of-the-art reasoning models, especially for hard problems
🧠 Excellent reasoning capabilities with high performance-to-cost ratio
Qwen/Qwen2-VL-7B-Instruct
Vision-language model that can process both images and text, enabling advanced multimodal reasoning capabilities
📸 Efficient image analysis and understanding at lower cost than proprietary alternatives
Qwen/Qwen2.5-Coder-32B-Instruct
Specialized model for code generation optimized for software development tasks
🔧 Superior in handling complex coding tasks, debugging, and documentation generation
Qwen/Qwen2.5-1.5B-Instruct
Latest series of Qwen large language models, efficient and powerful for general use
🚀 Ultra-efficient model for cost-sensitive deployments and edge applications
meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8
A 17 billion parameter multimodal model with 128 experts, beating GPT-4o and Gemini 2.0 Flash in benchmarks while using less than half the active parameters
🏆 Premium model combining best-in-class efficiency with multimodal capabilities
meta-llama/Llama-3.2-90B-Vision-Instruct
Multimodal model capable of processing images and text with 90B parameters, offering enhanced vision capabilities
👁️ Exceptional vision understanding for complex visual analysis and generation tasks
meta-llama/Llama-3.3-70B-Instruct
An auto-regressive language model using optimized transformer architecture with supervised fine-tuning (SFT) and RLHF for helpfulness and safety
🌐 Industry standard for high-quality general-purpose AI assistance and enterprise applications
mistralai/Mistral-Large-Instruct-2411
Advanced dense 7.3B parameter model with state-of-the-art reasoning, knowledge and coding capabilities with improved Long Context and Function Calling
💎 European-built model with exceptional parameter efficiency and GDPR considerations
mistralai/Ministral-8B-Instruct-2410
An instruct fine-tuned model significantly outperforming existing models of similar size, released under the Mistral Research License
⚖️ Optimal balance of compactness and capability for cost-effective production use
microsoft/phi-4
14B parameter state-of-the-art small language model (SLM) that excels at complex reasoning in math and conventional language processing
🧮 Exceptional at mathematical reasoning and analytical tasks despite compact size
microsoft/Phi-3.5-mini-instruct
Lightweight, state-of-the-art open model built upon high-quality, reasoning dense datasets with a focus on very high-quality reasoning tasks
📱 Ideal for mobile applications and edge computing with limited resources
databricks/dbrx-instruct
A mixture-of-experts (MoE) large language model trained from scratch by Databricks, specializing in few-turn interactions
📈 Particularly strong for data analysis workflows and integration with data platforms
netease-youdao/Confucius-o1-14B
O1-like reasoning model developed by NetEase Youdao Team, based on Qwen2.5-14B-Instruct with two-stage learning strategy for lightweight thinking abilities
🎯 Specialized in critical thinking and step-by-step problem solving
nvidia/AceMath-7B-Instruct
Models that excel at solving English mathematical problems using Chain-of-Thought (CoT) reasoning
📐 Specifically optimized for mathematical problem solving with step-by-step solutions
neuralmagic/Llama-3.1-Nemotron-70B-Instruct-HF-FP8-dynamic
Large language model customized by NV to improve the helpfulness of LLM generated responses to user queries
🔌 Optimized for hardware efficiency with dynamic quantization
SentientAGI/Dobby-Mini-Unhinged-Llama-3.1-8B
Language model fine-tuned from Llama-3.1-8B-Instruct with a strong conviction towards personal freedom, decentralization, and crypto
🪙 Specialized for crypto, blockchain, and Web3 applications
watt-ai/watt-tool-70B
Fine-tuned language model based on LLaMa-3.3-70B-Instruct, optimized for tool usage and multi-turn dialogue with state-of-the-art performance on the Berkeley Function-Calling Leaderboard
🔧 Exceptional at tool use, API interactions, and function calling
bespokelabs/Bespoke-Stratos-32B
Fine-tuned version of Qwen/Qwen2.5-32B-Instruct on the Bespoke-Stratos-17k dataset, derived by distilling DeepSeek-R1
🧩 Advanced problem-solving capabilities with specialized reasoning dataset
NovaSky-AI/Sky-T1-32B-Preview
32B reasoning model trained from Qwen2.5-32B-Instruct with 17K data, on par with o1-preview model on both math and coding
🌟 Balances capabilities across reasoning, mathematics, and coding tasks
tiiuae/Falcon3-10B-Instruct
State-of-the-art results on reasoning, language understanding, code and mathematics tasks, supporting 4 languages with a context length of up to 32K
🗣️ Strong multilingual capabilities with Middle Eastern language support
THUDM/glm-4-9b-chat
Open-source version of the latest generation of pre-trained models in the GLM-4 series launched by Zhipu AI
🀄 Excellent support for Asian languages, especially Chinese
CohereForAI/aya-expanse-32b
Open-weight research release of a model with highly advanced multilingual capabilities
🌍 Supports over 100 languages with strong cross-lingual transfer capabilities
jinaai/ReaderLM-v2
1.5B parameter model that converts raw HTML into beautifully formatted markdown or JSON with superior accuracy and longer context handling, supporting 29 languages
📚 Specialized in document processing with exceptional context length
openbmb/MiniCPM3-4B
4B parameter model with 32k context window, equipped with LLMxMapReduce to theoretically handle infinite context without requiring huge memory
📱 Ultra-lightweight model optimized for mobile and edge deployment
ozone-ai/0x-lite
State-of-the-art language model developed by Ozone AI, designed to deliver ultra-high-quality text generation while maintaining a compact architecture
✨ Optimized for natural language generation and creative writing tasks
ibm-granite/granite-3.1-8b-instruct
8B parameter long-context instruct model fine-tuned from Granite-3.1-8B-Base using a combination of open-source instruction datasets and internally collected synthetic datasets
🏢 Designed for enterprise use cases with business-focused training