Authentication
The following models are available through the Oraicle API:
meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8
A 17B-parameter Llama 4 MoE model, natively multimodal, 512K context, instruction-tuned, strong coding, multilingual, and image+text reasoning. 🦙
deepseek-ai/DeepSeek-R1-0528
Advanced open-source reasoning-optimized LLM, R1–0528, with unparalleled reasoning depth, long chain-of-thought, and high accuracy on complex tasks. 🧠
meta-llama/Llama-3.3-70B-Instruct
70B-parameter dense transformer, 128K context, instruction-tuned, strong reasoning, math, coding, and multilingual. 🌐
mistralai/Magistral-Small-2506
24B-parameter instruction-tuned model, 128K context, step-by-step reasoning, high multilingual capability. 🧑⚖️
mistralai/Devstral-Small-2505
24B-parameter coding/agentic LLM, 131K context, optimized for software engineering and tool use. 🛠️
mistralai/Mistral-Large-Instruct-2411
123B-parameter dense transformer, 128K context, SOTA reasoning, coding, and knowledge. 💎
Qwen/Qwen2.5-VL-32B-Instruct
32B-parameter vision-language, 8K+ context, RL-tuned, math/structured output, bilingual. 🖼️
meta-llama/Llama-3.2-90B-Vision-Instruct
90B-parameter multimodal, 128K context, vision+text, SOTA image reasoning, multilingual. 🦙
openai/gpt-oss-120b
OpenAI's open-weight 117B parameter Mixture-of-Experts model. 💥
openai/gpt-oss-20b
OpenAI's compact 20B open-weight and efficient AI model offering strong language processing power with a lighter footprint. ⚡️