https://mayoorrnikam.github.io/llm-world/ Every large language model release this project tracks, each traced to the lab's own announcement, paper, model card or documentation. en Thu, 15 Oct 2026 00:00:00 GMT https://www.rssboard.org/rss-specification https://mayoorrnikam.github.io/llm-world/models/nemotron-3-5-super-vl/ https://mayoorrnikam.github.io/llm-world/models/nemotron-3-5-super-vl/ Thu, 15 Oct 2026 00:00:00 GMT NVIDIA released Nemotron 3.5 Super VL on October 15, 2026. An open-weights multimodal mixture-of-experts model, 120B total and 12B active parameters, for text, image, document and video understanding with reasoning and tool use. NVIDIA says it was distilled from Nemotron 3 Ultra; base, post-trained and NVFP4 checkpoints were released. https://mayoorrnikam.github.io/llm-world/models/nemotron-3-5-bolt-74b-a9b/ https://mayoorrnikam.github.io/llm-world/models/nemotron-3-5-bolt-74b-a9b/ Thu, 15 Oct 2026 00:00:00 GMT NVIDIA released Nemotron 3.5 Bolt 74B-A9B on October 15, 2026. An open vision-language mixture-of-experts model, 74B total and 9B active parameters, built for local agentic work and packed into 24GB with 2-bit quantization-aware training. NVIDIA announced a smaller Bolt E4B alongside it for October 23. https://mayoorrnikam.github.io/llm-world/models/muse-spark-open/ https://mayoorrnikam.github.io/llm-world/models/muse-spark-open/ Thu, 15 Oct 2026 00:00:00 GMT Meta AI released Muse Spark Open on October 15, 2026. The open-weights version of Muse Spark, with additional safety post-training for open release: 816B total and 42B active parameters, natively multimodal, with a 1M-token context window. Weights are published by meta-models on Hugging Face under the Muse Model License, which limits commercial use. https://mayoorrnikam.github.io/llm-world/models/grok-voice-3-0/ https://mayoorrnikam.github.io/llm-world/models/grok-voice-3-0/ Wed, 14 Oct 2026 00:00:00 GMT xAI released Grok Voice 3.0 on October 14, 2026. xAI's speech-to-speech model, available in the API at launch at the same price as its predecessor. xAI says it already powers conversations across Tesla and SpaceX. https://mayoorrnikam.github.io/llm-world/models/muse-wonder/ https://mayoorrnikam.github.io/llm-world/models/muse-wonder/ Mon, 12 Oct 2026 00:00:00 GMT Meta AI released Muse Wonder on October 12, 2026. Meta Superintelligence Labs' first frontier model, announced with Muse Wonder Full as its most capable models, built for personal and professional agentic tasks. Available at announcement through the Meta Model API and Muse Code. No weights were published. https://mayoorrnikam.github.io/llm-world/models/muse-wonder-full/ https://mayoorrnikam.github.io/llm-world/models/muse-wonder-full/ Mon, 12 Oct 2026 00:00:00 GMT Meta AI released Muse Wonder Full on October 12, 2026. Announced alongside Muse Wonder as a separately named model. At announcement it was available only to Muse Code subscribers, not through the Meta Model API. Meta's posts do not say how it differs from Muse Wonder. https://mayoorrnikam.github.io/llm-world/models/microsoft-decision-1/ https://mayoorrnikam.github.io/llm-world/models/microsoft-decision-1/ Fri, 09 Oct 2026 00:00:00 GMT Microsoft released Microsoft-Decision-1 on October 9, 2026. A decision model from Microsoft's Office of the CTO, in public preview in Microsoft Foundry. Post-trained from Qwen3.5-9B for single-pass decision scoring: given yes/no, multiple-choice, rating or rubric options with a request, it returns a calibrated probability for each option rather than generated text. Microsoft says later versions will be rebased on OpenAI and MAI models. No weights were published. https://mayoorrnikam.github.io/llm-world/models/clef-omni/ https://mayoorrnikam.github.io/llm-world/models/clef-omni/ Fri, 09 Oct 2026 00:00:00 GMT Cloudflare released Clef-omni on October 9, 2026. A multimodal Clef that answers typed questions about audio, video, images and text together. Announced with a faster Clef and a cheaper Clef-flash. https://mayoorrnikam.github.io/llm-world/models/claude-haiku-5-5/ https://mayoorrnikam.github.io/llm-world/models/claude-haiku-5-5/ Wed, 07 Oct 2026 00:00:00 GMT Anthropic released Claude Haiku 5.5 on October 7, 2026. Anthropic's model tuned for high-volume, latency-sensitive work, released on the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry with a 1M-token context window, 128k max output tokens and adaptive thinking. Its API id is claude-haiku-5-5. https://mayoorrnikam.github.io/llm-world/models/d1-3b/ https://mayoorrnikam.github.io/llm-world/models/d1-3b/ Wed, 07 Oct 2026 00:00:00 GMT Liquid AI released d1-3B on October 7, 2026. An open-weights decision model built from LFM2.5-VL-3B. It takes text and images with typed questions supplied at request time (yes/no, choice, score) and answers in a single forward pass with probabilities and a confidence, generating no tokens. https://mayoorrnikam.github.io/llm-world/models/d1-omni-600m/ https://mayoorrnikam.github.io/llm-world/models/d1-omni-600m/ Wed, 07 Oct 2026 00:00:00 GMT Liquid AI released d1-omni-600M on October 7, 2026. The experimental small sibling of d1-3B, about 600M parameters, built from LFM2.5-Encoder-350M. It answers typed questions about text with images or audio in a single forward pass. https://mayoorrnikam.github.io/llm-world/models/nano-banana-2-1/ https://mayoorrnikam.github.io/llm-world/models/nano-banana-2-1/ Tue, 06 Oct 2026 00:00:00 GMT Google DeepMind released Nano Banana 2.1 on October 6, 2026. Google's high-efficiency image generation and conversational editing model, released generally available as gemini-nano-banana-2.1. Google describes it as an update to Nano Banana 2 (gemini-3.1-flash-image). https://mayoorrnikam.github.io/llm-world/models/embeddinggemma-2/ https://mayoorrnikam.github.io/llm-world/models/embeddinggemma-2/ Tue, 06 Oct 2026 00:00:00 GMT Google DeepMind released EmbeddingGemma 2 on October 6, 2026. Google's open-weights multimodal embedding model: 740M parameters mapping text, images, video frames and audio into one vector space, built to run on device. A text-only subset loads alone. https://mayoorrnikam.github.io/llm-world/models/mistral-large-4/ https://mayoorrnikam.github.io/llm-world/models/mistral-large-4/ Tue, 06 Oct 2026 00:00:00 GMT Mistral AI released Mistral Large 4 on October 6, 2026. A natively multimodal mixture-of-experts model with 1 trillion total and 52 billion active parameters, released as a public preview API on Mistral Studio. Mistral said it would release the weights by the end of the month; none were public at announcement. https://mayoorrnikam.github.io/llm-world/models/nova-2-5-sonic/ https://mayoorrnikam.github.io/llm-world/models/nova-2-5-sonic/ Mon, 05 Oct 2026 00:00:00 GMT Amazon released Nova 2.5 Sonic on October 5, 2026. Amazon's speech-to-speech model for real-time voice agents, generally available in Amazon Bedrock in four regions. Amazon describes improved reasoning, instruction following, tool calling and latency over Nova 2 Sonic, which this dataset does not yet record. https://mayoorrnikam.github.io/llm-world/models/mai-transcribe-2-streaming/ https://mayoorrnikam.github.io/llm-world/models/mai-transcribe-2-streaming/ Thu, 01 Oct 2026 00:00:00 GMT Microsoft released MAI-Transcribe-2-Streaming on October 1, 2026. Microsoft AI's first streaming transcription model: low-latency, real-time transcripts in 60 languages with continuous automatic language detection. Announced alongside MAI-Voice-2.1 and MAI-Voice-2.1-Flash. https://mayoorrnikam.github.io/llm-world/models/mai-voice-2-1/ https://mayoorrnikam.github.io/llm-world/models/mai-voice-2-1/ Thu, 01 Oct 2026 00:00:00 GMT Microsoft released MAI-Voice-2.1 on October 1, 2026. Microsoft AI's multilingual text-to-speech model, covering 23 languages and 26 locales. Announced alongside MAI-Transcribe-2-Streaming. https://mayoorrnikam.github.io/llm-world/models/mai-voice-2-1-flash/ https://mayoorrnikam.github.io/llm-world/models/mai-voice-2-1-flash/ Thu, 01 Oct 2026 00:00:00 GMT Microsoft released MAI-Voice-2.1-Flash on October 1, 2026. The lower-cost tier of MAI-Voice-2.1 for high-volume, latency-sensitive speech, with the same language coverage. Announced alongside MAI-Transcribe-2-Streaming. https://mayoorrnikam.github.io/llm-world/models/strands-decider-2b/ https://mayoorrnikam.github.io/llm-world/models/strands-decider-2b/ Thu, 01 Oct 2026 00:00:00 GMT Amazon released Strands Decider 2B on October 1, 2026. An open decision model from AWS's Strands Labs, citing TypeSafe AI's Jev as the model that drew attention to the class. Given a state and typed questions supplied with the request (a choice, a yes/no, a score), it returns each answer with a calibrated confidence from a small readout head, not by generating text. 1.9B parameters: a LoRA adapter on Qwen3.5-2B-Base plus that head, released with its training data and scripts. https://mayoorrnikam.github.io/llm-world/models/clef/ https://mayoorrnikam.github.io/llm-world/models/clef/ Thu, 01 Oct 2026 00:00:00 GMT Cloudflare released Clef on October 1, 2026. Cloudflare's open decision model, served on Workers AI. A frozen Qwen3.8-27B with low-rank adapters and a routing head: questions and answer types arrive with each request, and it scores the valid answers in one prefill pass, returning typed answers with probabilities rather than generating text.