AI model library
A proprietary reasoning-first Gemini model from Google, released 2026-02-19, accepting text, image, video, audio and PDF input with a 1M-token context window, aimed at agentic coding and complex problem solving.
GPT-5 Mini is OpenAI's small GPT-5 model with text and image input and a 400K-token context window, suited for responsive agents, coding help, and everyday automation.
OpenAI's large language model, part of the GPT family, specialized in dialogue and content creation.
A 0.8B open-weight vision-language model from Typhoon (SCB 10X), released under Apache-2.0, accepting text and image input for OCR-style image understanding and chat.
Typhoon OCR 2 (2B) is a 2.2B-parameter open-weight vision-language OCR model from SCB 10X's Typhoon team. It accepts text and image input for document/text recognition tasks and is released under Apache-2.0 for lightweight deployment.
d1-ultra is a model from Liquid AI's d1 family, released on 2026-10-16. It is closed-source with no open weights. Directory entry will be updated as more details emerge.
Reflection-Beam-1.0 is a model released on 2026-10-16. Its maker and specific capabilities cannot be verified from available tag data.
Intern-S2 is a large language model from Shanghai AI Laboratory, part of the Intern (书生) family, released in October 2026. Detailed capability and specification information has not been disclosed.
Ternary Bonsai 2 is a model released by PrismML on 2026-10-15 as part of the ternary-bonsai family. It is not open-weight, and no further capabilities or specifications are documented in the collected facts.
Grok Voice 3.0 is a voice model from xAI's Grok family, released in closed-source form on October 14, 2026.
Dyna-2 is a model from US robotics company Dyna Robotics, released on October 13, 2026. Its weights are not open, and no further capabilities or specifications are documented.
Natura-1 is a model released by Nature Will on 2026-10-13. It is not open-weights. Details on its capabilities and intended use cases are not documented in the collected facts.
GLM-Legal is a legal-domain model in the GLM family from Z.ai (Zhipu AI), released in October 2026. It is a closed-weights model aimed at legal use cases.
vLLM's decision model for decision support, released but not open-source.
Microsoft's decision model for enterprise decision optimization, proprietary.
Odyssey-3 is an AI model from the odyssey family, released on October 8, 2026, with no disclosed developer information and not open-sourced.
d1 is a model from US-based Liquid AI, released on 2026-10-07. It is not offered with open weights.
A 3B model in Liquid AI's d1 family, released in October 2026. Designed for lightweight deployment scenarios where a smaller model is preferred.
Rho-1 is a model from Reka AI's rho family, released in October 2026 with closed weights.
Strands Decider 2B is a closed-weight model from Amazon AWS, released in October 2026 as part of the Strands family, focused on decision-making tasks.
xAI's Grok Imagine Video 1.5 Lite is a video model for text-to-video and image-to-video generation.
Microsoft's MAI-Transcribe-2-Streaming, a real-time speech-to-text model for streaming audio transcription.
OpenAI's GPT series model focused on content summarization, scheduled for release on September 30, 2026.
TwIL-LM3-Pro is a closed-source LLM developed by webAI, released in September 2026.