https://llm-timeline.duyet.net Chronological index of Large Language Model releases from 2017 to present. Updated weekly. en-us Wed, 25 Mar 2026 00:00:00 GMT https://llm-timeline.duyet.net/favicon.svg https://llm-timeline.duyet.net https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Nemotron-Cascade-2-30B-A3B-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT NVIDIA open Parameters: Nemotron-Cascade-2-30B-A3B License: open Type: model 30BA3B. Gold medal performance in both the 2025 IMO and the IOI. HLE=no tools. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiMo-V2-Pro-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT Xiaomi open Parameters: MiMo-V2-Pro License: open Type: model 1T42B. Over 1T total parameters (42B active). Uses a 7:1 Hybrid Attention mechanism and supports a 1M-token context window. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Mamba-3-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT CMU open Parameters: Mamba-3 License: open Type: model "with architectural refinements, our Mamba-3 model achieves significant gains across retrieval, state-tracking, and downstream language modeling tasks." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiniMax-M2.5-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT MiniMax open Parameters: MiniMax-M2.5 License: open Type: model 230B-A10B. Early RSI. "M2.7 is our first model deeply participating in its own evolution…" https://lifearchitect.ai/asi/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Holotron-12B-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT H Company open Parameters: Holotron-12B License: open Type: model "Holotron-12B is a high-throughput, multimodal Vision-Language Model (VLM) designed specifically as a policy model for computer-use agents." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Mistral%20Small%204-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT Mistral open Parameters: Mistral Small 4 License: open Type: model 119BA6.5B. "unifies the capabilities of three different model families—Instruct, Reasoning (previously called Magistral), and Devstral—into a single, unified model." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiroThinker-H1-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT MiroMindAI open Parameters: MiroThinker-H1 License: open Type: model "Our proprietary agent, MiroThinker-H1 provides promising evidence for long-chain verifiable reasoning [based on new model, MiroThinker-1.7" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Covenant-72B-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT 1Covenant open Parameters: Covenant-72B License: open Type: model "largest permissionless collaboratively trained language model" ~20 distinct peers, each running 8xB200 GPUs. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Nemotron%203%20Super-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT NVIDIA open Parameters: Nemotron 3 Super License: open Type: model 120B-A12B. Announce: https://blogs.nvidia.com/blog/nemotron-3-super-agentic-ai/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Sarvam%20105B-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT Sarvam.ai open Parameters: Sarvam 105B License: open Type: model 105BA10.3B. "22 Indian languages" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.4-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT OpenAI open Parameters: GPT-5.4 License: open Type: model "most capable and efficient frontier model for professional work." Announce: https://openai.com/index/introducing-gpt-5-4/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Yuan3.0-Ultra-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT YuanLabAI open Parameters: Yuan3.0-Ultra License: open Type: model 1515BA68.8B. Poor performance due to low training data/ratio. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.3%20Instant-2026-03-01 Sun, 01 Mar 2026 00:00:00 GMT OpenAI open Parameters: GPT-5.3 Instant License: open Type: model Announce: https://openai.com/index/gpt-5-3-instant/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#STATIC-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Google open Parameters: STATIC License: open Type: model YouTube (Google). STATIC (Sparse Transition Matrix-Accelerated Trie Index for Constrained Decoding). "The model is a Gemini-based generative retrieval model similar to PLUM [8], served with a batch size of 2 (per chip) and a beam size of 𝑀 = 70. The model is based on a non-Mixture-of-Experts (MoE) architecture with 3 billion dense parameters. All benchmark experiments are conducted on Google TPU v6e accelerators." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Arrow%201.0-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Quiver open Parameters: Arrow 1.0 License: open Type: model "A first of it's kind SVG AI model." Announce: https://x.com/QuiverAI/status/2026792057893708072 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Qwen3.5-27B-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Alibaba open Parameters: Qwen3.5-27B License: open Type: model "Qwen3.5 represents a significant leap forward, integrating breakthroughs in multimodal learning, architectural efficiency, reinforcement learning scale, and global accessibility" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#LFM2-24B-A2B-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Liquid AI open Parameters: LFM2-24B-A2B License: open Type: model "a traditional instruct model without reasoning traces." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Mercury%202-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Inception open Parameters: Mercury 2 License: open Type: model Diffusion large language model (dLLM). https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Gemini%203.1%20Pro-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Google DeepMind open Parameters: Gemini 3.1 Pro License: open Type: model Knowledge cutoff still=January 2025. Announce: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#ZUNA-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Zyphra open Parameters: ZUNA License: open Type: model For BCI, 'thought-to-text'. Training dataset calcs: (2M hours * 3,600 seconds/hour * 256 samples/second ) / 32 samples/token = 57.6B tokens (refined to 45.1B after rigorous filtering ); 150,000 steps * 2.16M tokens/batch = 324B total tokens seen during training. Announce: https://www.zyphra.com/post/zuna https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Grok%204.2-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT xAI open Parameters: Grok 4.2 License: open Type: model No details provided. Announce: https://x.com/elonmusk/status/2023829664318583105 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#INTELLECT-3.1-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Prime Intellect open Parameters: INTELLECT-3.1 License: open Type: model Base: GLM-4.5-Air-Base, INTELLECT-3 model. 106BA12B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Claude%20Sonnet%204.6-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Anthropic open Parameters: Claude Sonnet 4.6 License: open Type: model 1M context. Announce: https://www.anthropic.com/news/claude-sonnet-4-6 Showing GMMLU (Global MMLU by Cohere). https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Tiny%20Aya-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Cohere open Parameters: Tiny Aya License: open Type: model 70+ languages. Showing GMMLU (Global MMLU by Cohere). https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Qwen3.5-397B-A17B-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Alibaba open Parameters: Qwen3.5-397B-A17B License: open Type: model "Qwen3.5 represents a significant leap forward, integrating breakthroughs in multimodal learning, architectural efficiency, reinforcement learning scale, and global accessibility" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#JoyAI-LLM%20Flash-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT JD Open Source open Parameters: JoyAI-LLM Flash License: open Type: model 48B-A3B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiniMax-M2.5-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT MiniMax open Parameters: MiniMax-M2.5 License: open Type: model 230B-A10B. HLE showing without tools. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GLM-5-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Z.AI open Parameters: GLM-5 License: open Type: model 744B-A40B. Announce: https://z.ai/blog/glm-5 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Nanbeige4.1-3B-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Nanbeige open Parameters: Nanbeige4.1-3B License: open Type: model SOTA for size (3B) https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#RynnBrain-30B-A3B-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Alibaba open Parameters: RynnBrain-30B-A3B License: open Type: model Base: Qwen3-VL-30B-A3B-Instruct. "an embodied foundation model grounded in physical reality." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Claude%20Opus%204.6-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Anthropic open Parameters: Claude Opus 4.6 License: open Type: model https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Intern-S1-Pro-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT Shanghai AI Laboratory/SenseTime open Parameters: Intern-S1-Pro License: open Type: model 1000TA22B. Assumes base model of Qwen3. "Built upon a 235B MoE language model and a 6B Vision encoder, Intern-S1 has been further pretrained on 5 trillion tokens of multimodal data" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Step%203.5%20Flash-2026-02-01 Sun, 01 Feb 2026 00:00:00 GMT StepFun open Parameters: Step 3.5 Flash License: open Type: model 196B-A11B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Assistant_Pepe_8B-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Independent open Parameters: Assistant_Pepe_8B License: open Type: model Warning for inappropriate content. Base: Llama-3.1-Nemotron-8B. "trained it on an extended 4chan dataset" "the original, gpt4chan (by Yannic Kilcher) scored especially high in truthfulness (that was b4 benchmaxxing)... outperformed the base tune (the unabliterated one), it also changed its political alignment... People were initially joking about the "alignment tax", I think there's a none trivial substance in all of this. It seems to me just above a marginal error or statistical noise." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Trinity-Large-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Arcee AI open Parameters: Trinity-Large License: open Type: model 400BA13B. "we worked closely with Prime Intellect. They not only served the H100 clusters Datology used to generate synthetic data, they have been deeply involved in helping scale our training setup to the GPU footprint required for a fully frontier sized model, including the current 2048 B300 GPU configuration for Trinity Large." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#SERA-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Allen AI open Parameters: SERA License: open Type: model Base: Qwen3-32B. SERA=Soft-verified Efficient Repository Agents. "SERA was built largely by a single Ai2 researcher." https://allenai.org/blog/open-coding-agents "SERA-32B was trained using Soft Verified Generation (SVG), a simple and efficient method that is 26x cheaper than reinforcement learning and 57x cheaper than previous synthetic data methods to reach equivalent performance. The total cost for data generation and training is approximately $2,000 (40 GPU-days)." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Kimi%20K2.5-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Moonshot AI open Parameters: Kimi K2.5 License: open Type: model 1TA32B. 1T parameters and 384 experts. Open source SOTA. "Kimi K2.5 builds on Kimi K2 [15.5T tokens] with continued pretraining over approximately 15T mixed visual and text tokens. [+ 15T=30.5T]" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GLM-4.7-Flash-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Z.AI open Parameters: GLM-4.7-Flash License: open Type: model 30B-A3B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MedGemma%201.5%204B-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Google DeepMind open Parameters: MedGemma 1.5 4B License: open Type: model Lower MMLU score compared to previous MedGemma 1 27B (67.2 v 87). Announce: https://research.google/blog/next-generation-medical-image-interpretation-with-medgemma-15-and-medical-speech-to-text-with-medasr/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#FrogBoss-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Microsoft open Parameters: FrogBoss License: open Type: model Base: Qwen3-32B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#EDEN-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT NVIDIA closed Parameters: EDEN License: closed Type: model "EDEN (environmentally-derived evolutionary network) family of metagenomic foundation models, including a 28 billion parameter model trained on 9.7 trillion nucleotide tokens from BaseData1 . This dataset, at the time of training, contained more than 10 billion novel genes from over 1 million new species, and is intentionally enriched for environmental and host-associated metagenomes, phage sequences, and mobile genetic elements, enabling the model to learn from diverse and novel cross-species evolutionary mechanisms and apply them to key challenges in human health." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Baichuan-M3-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Baichuan open Parameters: Baichuan-M3 License: open Type: model "new-generation medical-enhanced large language model" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Engram-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT DeepSeek-AI partial Parameters: Engram License: partial Type: model 39.5BA3.8B. "we explore conditional memory as a complementary sparsity axis, instantiated via Engram, a module that modernizes classic N -gram embeddings for O ( 1 ) lookup." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#SleepFM-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Stanford open Parameters: SleepFM License: open Type: model Uses a leave-one-out contrastive learning approach to align brain activity (EEG), heart activity (ECG), and respiratory signals. 130+ disease categories and 19–20+ clinical PSG channels. Dataset ~12.63B (Calculated based on 585,000 hours of data across 3 modality groups using 5-second window tokens) x 10 epochs. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#TimeCapsuleLLM-v2-1800-1875-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Independent open Parameters: TimeCapsuleLLM-v2-1800-1875 License: open Type: model 112GB dataset=30B tokens x 0.5 epochs = 15B tokens. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Jamba2-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT AI21 open Parameters: Jamba2 License: open Type: model 52B-A12B. Pre-training tokens from Jamba=1.2T + 500B mid. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#LFM2.5-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT Liquid AI open Parameters: LFM2.5 License: open Type: model For on-device agentic applications. "Extended pre-training from 10T to 28T tokens and large-scale multi-stage reinforcement learning." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiroThinker%20v1.5-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT MiroMindAI open Parameters: MiroThinker v1.5 License: open Type: model Base: Qwen3 235B-A22B. Official demo: https://dr.miromind.ai https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Falcon-H1R-2026-01-01 Thu, 01 Jan 2026 00:00:00 GMT TII open Parameters: Falcon-H1R License: open Type: model Base model: Falcon-H1 (May/2025). Announce: https://huggingface.co/blog/tiiuae/falcon-h1r-7b https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Solar%20Open%20100B-2025-12-31 Wed, 31 Dec 2025 00:00:00 GMT Upstage closed Parameters: 102B License: closed Type: model AI model by Upstage https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#K-EXAONE-2025-12-31 Wed, 31 Dec 2025 00:00:00 GMT LG AI Research closed Parameters: 236B License: closed Type: model AI model by LG AI Research https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#VAETKI-2025-12-30 Tue, 30 Dec 2025 00:00:00 GMT NC AI open Parameters: 100B License: open Type: model AI model by NC AI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#A.X%20K1-2025-12-30 Tue, 30 Dec 2025 00:00:00 GMT SK Telecom closed Parameters: 519B License: closed Type: model AI model by SK Telecom https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#HyperCLOVA%20X%20SEED%2032B%20Think-2025-12-29 Mon, 29 Dec 2025 00:00:00 GMT NAVER closed Parameters: 32B License: closed Type: model AI model by NAVER https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiniMax-M2.1-2025-12-23 Tue, 23 Dec 2025 00:00:00 GMT MiniMax open Parameters: 229B License: open Type: model AI model by MiniMax https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GLM-4.7-2025-12-22 Mon, 22 Dec 2025 00:00:00 GMT Z.ai (Zhipu AI) open Parameters: 358B License: open Type: model AI model by Z.ai (Zhipu AI) https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.2%20Codex-2025-12-18 Thu, 18 Dec 2025 00:00:00 GMT OpenAI closed License: closed Type: model AI model by OpenAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Nomos%201-2025-12-11 Thu, 11 Dec 2025 00:00:00 GMT Nous Research open Parameters: 30B License: open Type: model AI model by Nous Research https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Nova%202-2025-12-02 Tue, 02 Dec 2025 00:00:00 GMT Amazon Web Services (AWS) closed License: closed Type: model AI model by Amazon Web Services (AWS) https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#mHC%2027B-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT DeepSeek-AI closed Parameters: mHC 27B License: closed Type: model 27BA4.14B. Scaling tested with 3B MoE on 1T tokens=334:1. "Manifold-Constrained Hyper-Connections (mHC), a general framework that projects the residual connection space of HC onto a specific manifold to restore the identity mapping property, while incorporating rigorous infrastructure optimization to ensure efficiency. Empirical experiments demonstrate that mHC is effective for training at scale, offering tangible performance improvements and superior scalability." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#IQuest-Coder-V1-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT IQuestLab open Parameters: IQuest-Coder-V1 License: open Type: model "IQuest-Coder-V1 captures the dynamic evolution of software logic, delivering state-of-the-art performance across critical dimensions" https://github.com/IQuestLab/IQuest-Coder-V1 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#A.X%20K1-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT SK Hynix open Parameters: A.X K1 License: open Type: model 519BA33B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#K-EXAONE-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT LG open Parameters: K-EXAONE License: open Type: model 236BA23B. “EXAONE”=“EXpert AI for EveryONE”. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Ranke-4B-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT UZH closed Parameters: Ranke-4B License: closed Type: model Base Model: Qwen 3. 600B tokens of pre-(1913, 1929, 1933, 1939, 1946) data only. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#WeDLM-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Tencent open Parameters: WeDLM License: open Type: model Project page: https://wedlm.github.io/ "WeDLM, a diffusion decoding framework built entirely on standard causal attention to make parallel generation prefix-cache friendly. The core idea is to let each masked position condition on all currently observed tokens while keeping a strict causal mask, achieved by Topological Reordering that moves observed tokens to the physical prefix while preserving their logical positions.. We instantiate WeDLM on both Qwen2.5-7B and Qwen3-8B, utilizing 100B tokens for continued training and 10B tokens for SFT." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#SOLAR%20Open-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Upstage AI open Parameters: SOLAR Open License: open Type: model South Korean. 102BA12B. Releasing 31/Dec. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GLM-4.7-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Z.AI open Parameters: GLM-4.7 License: open Type: model 355B-A32B. "context window has been expanded from 128K to 200K tokens" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#NitroGen-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT NVIDIA open Parameters: NitroGen License: open Type: model "NitroGen is a unified vision-to-action model designed to play video games directly from raw frames. It takes video game footage as input and outputs gamepad actions... trained on 40,000 hours of gameplay videos across more than 1,000 games." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiMo-V2-Flash-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Xiaomi open Parameters: MiMo-V2-Flash License: open Type: model 309BA15B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#FunctionGemma-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Google DeepMind open Parameters: FunctionGemma License: open Type: model "FunctionGemma, a specialized version of our Gemma 3 270M model tuned for function calling. It is designed as a strong base for further training into custom, fast, private, local agents that translate natural language into executable API actions." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#T5Gemma%202-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Google DeepMind open Parameters: T5Gemma 2 License: open Type: model Base model: Gemma 3. Dataset: Gemma 3 4B checkpoint (4T) + pretraining (2T)=6T. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Gemini%203%20Flash-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Google DeepMind open Parameters: Gemini 3 Flash License: open Type: model Announce: https://deepmind.google/models/gemini/flash/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#NVIDIA-Nemotron-3-Nano-30B-A3B-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT NVIDIA open Parameters: NVIDIA-Nemotron-3-Nano-30B-A3B License: open Type: model Knowledge cutoff November 28, 2025 (post). https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Bolmo-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Allen AI open Parameters: Bolmo License: open Type: model Base Model: Olmo 3 7B. Announce: https://allenai.org/blog/bolmo https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#EuroLLM-22B-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Consortium open Parameters: EuroLLM-22B License: open Type: model A fully open language model developed in Europe. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#LLaDA2.0%20Flash-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Inclusion AI open Parameters: LLaDA2.0 Flash License: open Type: model Base Model: Ling-flash-2.0: 103B total parameters with 6.1B activated. "largest diffusion language model to date" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.2-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT OpenAI open Parameters: GPT-5.2 License: open Type: model "GPT‑5.2 sets a new state of the art across many benchmarks, including GDPval, where it outperforms industry professionals at well-specified knowledge work tasks spanning 44 occupations." Announce: https://openai.com/index/introducing-gpt-5-2/ MMLU is for Spanish. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Apriel-1.6-15B-Thinker-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT ServiceNow open Parameters: Apriel-1.6-15B-Thinker License: open Type: model https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Motif%202%2012.7B-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Motif-Technologies open Parameters: Motif 2 12.7B License: open Type: model https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Devstral%202-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Mistral open Parameters: Devstral 2 License: open Type: model SWE-bench Verified=72.2%. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Nanbeige4-3B-Base-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Nanbeige4-3B-Base open Parameters: Nanbeige4-3B-Base License: open Type: model https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#HY%202.0-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Tencent open Parameters: HY 2.0 License: open Type: model 406BA32B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#K2-V2-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT MBZUAI open Parameters: K2-V2 License: open Type: model 8.5x more tokens trained than K2 (1.4T v 12T). Project page: https://ifm.ai/k2/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Trinity-Mini-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Arcee AI open Parameters: Trinity-Mini License: open Type: model 26BA3B. "we worked closely with Prime Intellect. They not only served the H100 clusters Datology used to generate synthetic data, they have been deeply involved in helping scale our training setup to the GPU footprint required for a fully frontier sized model, including the current 2048 B300 GPU configuration for Trinity Large." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Nova%202%20Pro-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Amazon open Parameters: Nova 2 Pro License: open Type: model "Nova 2 Pro is Amazon's most intelligent reasoning model that can process text, images, video, and speech to generate text." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Mistral%20Large%203-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT Mistral open Parameters: Mistral Large 3 License: open Type: model 675BA41B. "Mistral Large 3 joins the ranks of frontier instruction-fine-tuned open-source models." EU tech doc: https://legal.cms.mistral.ai/assets/1e37fffd-7ea5-469b-822f-05dcfbb43623 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#DeepSeek-V3.2-Speciale-2025-12-01 Mon, 01 Dec 2025 00:00:00 GMT DeepSeek-AI open Parameters: DeepSeek-V3.2-Speciale License: open Type: model The word 'Speciale' may be a reference to Ferrari. "It shows gold-medal performance in the IOI 2025, ICPC World Final 2025, IMO 2025, and CMO 2025." API: https://api-docs.deepseek.com/news/news251201 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#DeepSeekMath-V2-2025-11-27 Thu, 27 Nov 2025 00:00:00 GMT DeepSeek open Parameters: 685B License: open Type: model AI model by DeepSeek https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Claude%20Opus%204.5-2025-11-24 Mon, 24 Nov 2025 00:00:00 GMT Anthropic closed License: closed Type: model AI model by Anthropic https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Olmo%203-2025-11-20 Thu, 20 Nov 2025 00:00:00 GMT Allen Institute for AI (Ai2) open Parameters: 32B License: open Type: model AI model by Allen Institute for AI (Ai2) https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Grok%204.1%20Fast-2025-11-19 Wed, 19 Nov 2025 00:00:00 GMT xAI closed License: closed Type: model AI model by xAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.1-Codex-Max-2025-11-19 Wed, 19 Nov 2025 00:00:00 GMT OpenAI closed License: closed Type: model AI model by OpenAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Gemini%203%20Pro-2025-11-18 Tue, 18 Nov 2025 00:00:00 GMT Google DeepMind closed License: closed Type: model AI model by Google DeepMind https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Grok%204.1-2025-11-17 Mon, 17 Nov 2025 00:00:00 GMT xAI closed License: closed Type: model AI model by xAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.1-2025-11-13 Thu, 13 Nov 2025 00:00:00 GMT OpenAI closed License: closed Type: model AI model by OpenAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.1%20Instant-2025-11-13 Thu, 13 Nov 2025 00:00:00 GMT OpenAI closed License: closed Type: model AI model by OpenAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.1-Codex-2025-11-12 Wed, 12 Nov 2025 00:00:00 GMT OpenAI closed License: closed Type: model AI model by OpenAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Kimi%20K2%20Thinking-2025-11-06 Thu, 06 Nov 2025 00:00:00 GMT Moonshot open Parameters: 1T License: open Type: model AI model by Moonshot https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Gen-0-2025-11-04 Tue, 04 Nov 2025 00:00:00 GMT Generalist closed Parameters: 10B License: closed Type: model AI model by Generalist https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#DeepSeek-Math-V2-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT DeepSeek-AI open Parameters: DeepSeek-Math-V2 License: open Type: model "DeepSeekMath-V2, demonstrates strong theorem-proving capabilities, achieving gold-level scores on IMO 2025 and CMO 2024 and a near-perfect 118/120 on Putnam 2024 with scaled testtime compute. " https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Orchestrator-8B-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT NVIDIA open Parameters: Orchestrator-8B License: open Type: model Base Model: Qwen3-8B https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#INTELLECT-3-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Prime Intellect open Parameters: INTELLECT-3 License: open Type: model Base: GLM-4.5-Air-Base model. 106BA12B. Announce: https://www.primeintellect.ai/blog/intellect-3 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Fara-7B-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Microsoft open Parameters: Fara-7B License: open Type: model "Fara-7B is Microsoft's first agentic small language model (SLM) designed specifically for computer use. With only 7 billion parameters, Fara-7B is an ultra-compact Computer Use Agent (CUA)...Current production baselines leverage Qwen 2.5-VL (7B)." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Claude%20Opus%204.5-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Anthropic open Parameters: Claude Opus 4.5 License: open Type: model "the best model in the world for coding, agents, and computer use." Announce: https://www.anthropic.com/news/claude-opus-4-5 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Nemotron%20Elastic-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT NVIDIA open Parameters: Nemotron Elastic License: open Type: model "Nemotron Elastic, a framework for building reasoning-oriented LLMs, including hybrid Mamba-Attention architectures, that embed multiple nested submodels within a single parent model, each optimized for different deployment configurations and budgets. Each of these submodels shares weights with the parent model and can be extracted zero-shot during deployment without additional training or fine-tuning...We apply Nemotron Elastic to the Nemotron Nano V2 12B model, simultaneously producing a 9B and a 6B model using only 110B training tokens" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GeoVista-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Tencent open Parameters: GeoVista License: open Type: model Base model: Qwen2.5-VL-7B-Instruct. "GeoVista, an agentic model that seamlessly integrates tool invocation within the reasoning loop, including an image-zoom-in tool to magnify regions of interest and a web-search tool to retrieve related web information. " Project page: https://ekonwang.github.io/geo-vista/ https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#OLMo%203-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Allen AI open Parameters: OLMo 3 License: open Type: model Announce: https://allenai.org/blog/olmo3 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Gemini%203%20Pro-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Google DeepMind open Parameters: Gemini 3 Pro License: open Type: model "The knowledge cutoff date for Gemini 3 Pro was January 2025." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Grok%204.1-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT xAI open Parameters: Grok 4.1 License: open Type: model https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Baguettotron-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT PleIAs open Parameters: Baguettotron License: open Type: model "The name is both a nod to French origins and to the unusual shape of the model: with 80 layers, Baguettotron is currently the deepest SLM in its size range." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#ERNIE-5.0-Preview-1022-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Baidu open Parameters: ERNIE-5.0-Preview-1022 License: open Type: model Very low performance on ALPrompt. 2.4T params confirmed: https://global.chinadaily.com.cn/a/202511/13/WS691571bda310d6866eb29500.html https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5.1-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT OpenAI open Parameters: GPT-5.1 License: open Type: model Personality change via fine-tuning. GPQA (no tools) increased from GPT-5=85.7 to GPT-5.1=88.1. MMLU is for Spanish. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#TiDAR-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT NVIDIA open Parameters: TiDAR License: open Type: model Base model: Qwen3-8B (36T) + 150B continual training. "TiDAR, a sequence-level hybrid architecture that drafts tokens (Thinking) in Diffusion and samples final outputs (Talking) AutoRegressively - all within a single forward pass using specially designed structured attention masks" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#SONIC-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT NVIDIA open Parameters: SONIC License: open Type: model Supersizing mOtion tracking for Natural humanoId Control (SONIC). Training dataset calcs: (700 hours * 3,600 seconds/hour * 50 frames/second ) / 1 frame/token = 126M tokens (refined to 100M+ after rigorous filtering ); 150,000 steps * 6.67M tokens/batch = 1.0T total tokens seen during training. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#JustRL-Nemotron-1.5B-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Tsinghua open Parameters: JustRL-Nemotron-1.5B License: open Type: model "JustRL, a simple recipe with fixed hyperparameters, achieves state-of-the-art performance on two different 1.5B base models (54.5% and 64.3% across 9 math benchmarks) while using 2× less compute than sophisticated approaches. The same hyperparameters transfer across both models without tuning, and training remains stable over thousands of steps without intervention. This suggests the field may be adding complexity to solve problems that disappear with a stable, scaled-up baseline." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#ERNIE-4.5-VL-28B-A3B-Thinking-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Baidu open Parameters: ERNIE-4.5-VL-28B-A3B-Thinking License: open Type: model 28B-A3B. Open-sourced 12/Nov/2025 from Jun/2025 release. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#HOPE-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Google DeepMind partial Parameters: HOPE License: partial Type: model "Combining our self-modifying sequence model with the continuum memory system, we present a learning module, called HOPE, showing promising results in language modeling, continual learning, and long-context reasoning tasks." Announce: https://research.google/blog/introducing-nested-learning-a-new-ml-paradigm-for-continual-learning/ May be released after paper is public. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Kimi%20K2%20Thinking-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Moonshot AI open Parameters: Kimi K2 Thinking License: open Type: model 1TA32B. 1T parameters and 384 experts. Open source SOTA. HLE=51.0 on text-only subset, compare to Grok-4 HLE=50.7 also on text-only, but Grok-4 HLE=44.4 on HLE full, ∴ Kimi K2 Thinking HLE≈44 full (estimated). https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Ling-1T-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Inclusion AI open Parameters: Ling-1T License: open Type: model 1TA50B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GEN-0-2025-11-01 Sat, 01 Nov 2025 00:00:00 GMT Generalist partial Parameters: GEN-0 License: partial Type: model "GEN-0, a new class of embodied foundation models built for multimodal training directly on high-fidelity raw physical interaction. Its architecture builds on the strengths of vision and language models while also going beyond them—natively designed to capture human-level reflexes and physical commonsense. One core feature is Harmonic Reasoning, in which the models are trained to simultaneously think and act seamlessly... GEN-0 is pretrained on our in-house robotics dataset, which includes over 270,000 hours of real-world diverse manipulation data, growing at a rate of 10,000 hours a week and accelerating." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Emu3.5-2025-10-30 Thu, 30 Oct 2025 00:00:00 GMT Beijing Academy of Artificial Intelligence / BAAI open Parameters: 34.1B License: open Type: model AI model by Beijing Academy of Artificial Intelligence / BAAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Kimi%20Linear-2025-10-30 Thu, 30 Oct 2025 00:00:00 GMT Moonshot open Parameters: 48B License: open Type: model AI model by Moonshot https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Composer-2025-10-29 Wed, 29 Oct 2025 00:00:00 GMT Cursor closed License: closed Type: model AI model by Cursor https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#SWE-1.5-2025-10-29 Wed, 29 Oct 2025 00:00:00 GMT Cognition closed Parameters: 300B License: closed Type: model AI model by Cognition https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Tongyi%20DeepResearch-2025-10-28 Tue, 28 Oct 2025 00:00:00 GMT Alibaba open Parameters: 30.5B License: open Type: model AI model by Alibaba https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiniMax-M2-2025-10-27 Mon, 27 Oct 2025 00:00:00 GMT MiniMax open Parameters: 229B License: open Type: model AI model by MiniMax https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#LoongRL%207B-2025-10-27 Mon, 27 Oct 2025 00:00:00 GMT Microsoft Research Asia,Shanghai Jiao Tong University,Carnegie Mellon University (CMU) closed Parameters: 7B License: closed Type: model AI model by Microsoft Research Asia,Shanghai Jiao Tong University,Carnegie Mellon University (CMU) https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#LoongRL%2014B-2025-10-27 Mon, 27 Oct 2025 00:00:00 GMT Microsoft Research Asia,Shanghai Jiao Tong University,Carnegie Mellon University (CMU) closed Parameters: 14B License: closed Type: model AI model by Microsoft Research Asia,Shanghai Jiao Tong University,Carnegie Mellon University (CMU) https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Lapa%20LLM-2025-10-25 Sat, 25 Oct 2025 00:00:00 GMT Ukrainian Catholic University,Igor Sikorsky Kyiv Polytechnic Institute,AGH University of Krakow,Lviv Polytechnic open Parameters: 12B License: open Type: model AI model by Ukrainian Catholic University,Igor Sikorsky Kyiv Polytechnic Institute,AGH University of Krakow,Lviv Polytechnic https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Ring-mini-linear-2.0-2025-10-23 Thu, 23 Oct 2025 00:00:00 GMT Ant Group open Parameters: 16.4B License: open Type: model AI model by Ant Group https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Ring-flash-linear-2.0-2025-10-23 Thu, 23 Oct 2025 00:00:00 GMT Ant Group open Parameters: 104.2B License: open Type: model AI model by Ant Group https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Deepseek%20OCR-2025-10-21 Tue, 21 Oct 2025 00:00:00 GMT DeepSeek open Parameters: 3B License: open Type: model AI model by DeepSeek https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#BAPO%2032B-2025-10-21 Tue, 21 Oct 2025 00:00:00 GMT Fudan University,Shanghai Qiji Zhifeng closed Parameters: 32B License: closed Type: model AI model by Fudan University,Shanghai Qiji Zhifeng https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Odyssey%20102B-2025-10-18 Sat, 18 Oct 2025 00:00:00 GMT Anthrogen closed Parameters: 102B License: closed Type: model AI model by Anthrogen https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Odyssey%2012B-2025-10-18 Sat, 18 Oct 2025 00:00:00 GMT Anthrogen closed Parameters: 12B License: closed Type: model AI model by Anthrogen https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Odyssey%201.2B-2025-10-18 Sat, 18 Oct 2025 00:00:00 GMT Anthrogen closed Parameters: 1.2B License: closed Type: model AI model by Anthrogen https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Claude%20Haiku%204.5-2025-10-15 Wed, 15 Oct 2025 00:00:00 GMT Anthropic closed License: closed Type: model AI model by Anthropic https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Veo%203.1-2025-10-15 Wed, 15 Oct 2025 00:00:00 GMT Google DeepMind closed License: closed Type: model AI model by Google DeepMind https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Llama%204%20Scout%20%2B%20ScaleRL-2025-10-15 Wed, 15 Oct 2025 00:00:00 GMT Meta AI closed Parameters: 109B License: closed Type: model AI model by Meta AI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MAI-Image-1-2025-10-13 Mon, 13 Oct 2025 00:00:00 GMT Microsoft closed License: closed Type: model AI model by Microsoft https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#C2S-Scale-2025-10-11 Sat, 11 Oct 2025 00:00:00 GMT Google Research,Yale University,Google DeepMind,Brown University,University of Southern California open Parameters: 27B License: open Type: model AI model by Google Research,Yale University,Google DeepMind,Brown University,University of Southern California https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Ring-1T-2025-10-10 Fri, 10 Oct 2025 00:00:00 GMT Ant Group open Parameters: 1T License: open Type: model AI model by Ant Group https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Ling-1T-2025-10-10 Fri, 10 Oct 2025 00:00:00 GMT Ant Group open Parameters: 1T License: open Type: model AI model by Ant Group https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Grok%20Imagine-2025-10-08 Wed, 08 Oct 2025 00:00:00 GMT xAI closed License: closed Type: model AI model by xAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT-5%20Pro-2025-10-07 Tue, 07 Oct 2025 00:00:00 GMT OpenAI closed License: closed Type: model AI model by OpenAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Gemini%202.5%20Computer%20Use-2025-10-07 Tue, 07 Oct 2025 00:00:00 GMT Google closed License: closed Type: model AI model by Google https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Tiny%20Recursive%20Model%20(TRM-Att)-2025-10-06 Mon, 06 Oct 2025 00:00:00 GMT Samsung SAIT AI Lab closed Parameters: 7M License: closed Type: model AI model by Samsung SAIT AI Lab https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Granite-4.0-H-Tiny-2025-10-02 Thu, 02 Oct 2025 00:00:00 GMT IBM open Parameters: 7B License: open Type: model AI model by IBM https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Granite-4.0-H-Micro-2025-10-02 Thu, 02 Oct 2025 00:00:00 GMT IBM open Parameters: 3B License: open Type: model AI model by IBM https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Granite-4.0-H-Small-2025-10-02 Thu, 02 Oct 2025 00:00:00 GMT IBM open Parameters: 32B License: open Type: model AI model by IBM https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#CALM-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT Wechat open Parameters: CALM License: open Type: model "Continuous Autoregressive Language Models (CALM), a paradigm shift from discrete next-token prediction to continuous next-vector prediction. CALM uses a high-fidelity autoencoder to compress a chunk of K tokens into a single continuous vector, from which the original tokens can be reconstructed with over 99.9% accuracy... We train our models on the Pile uncopyrighted dataset (Gao et al., 2020). The raw text is processed with the Llama 3 tokenizer (Grattafiori et al., 2024), resulting in a training set of ∼230B tokens." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Kimi-Linear-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT Moonshot AI open Parameters: Kimi-Linear License: open Type: model 48B-A3B. "Kimi Linear is a hybrid linear attention architecture that outperforms traditional full attention methods across various contexts, including short, long, and reinforcement learning (RL) scaling regimes. At its core is Kimi Delta Attention (KDA)—a refined version of Gated DeltaNet that introduces a more efficient gating mechanism to optimize the use of finite-state RNN memory." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MiniMax-M2-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT MiniMax open Parameters: MiniMax-M2 License: open Type: model 230B-A10B. https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MACE-MH-1-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT Cambridge/LBNL open Parameters: MACE-MH-1 License: open Type: model MACE-MH-1 (Multi-Head 1). Features Multiple Heads (OMAT PBE, OMOL r2scan, OC20) to maintain high accuracy across domains https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#DeepSeek-OCR-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT DeepSeek-AI open Parameters: DeepSeek-OCR License: open Type: model 2D vision tokens for 1D text achieves huge compression. Encoder/Decoder: DeepEncoder 380M (80M SAM-base + 300M CLIP-large), DeepSeek-3B-MoE (A570M). https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#UserLM-8b-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT Microsoft open Parameters: UserLM-8b License: open Type: model "we trained UserLM-8b to simulate the “user” role in conversation (by training it to predict user turns in a large corpus of conversations called WildChat)." https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#CoDA-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT Salesforce open Parameters: CoDA License: open Type: model "diffusion coder trained on TPU [Google TPU v4-1024 VM]" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#TRM-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT Samsung open Parameters: TRM License: open Type: model "Tiny Recursive Model (TRM), a much simpler recursive reasoning approach that achieves significantly higher generalization than HRM, while using a single tiny network with only 2 layers" https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Granite-4.0%20Small-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT IBM open Parameters: Granite-4.0 Small License: open Type: model 32B-A9B. Announce: https://www.ibm.com/new/announcements/ibm-granite-4-0-hyper-efficient-high-performance-hybrid-models https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Octave%202-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT Hume closed License: closed Type: model AI model by Hume https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#EVI%204%20mini-2025-10-01 Wed, 01 Oct 2025 00:00:00 GMT Hume closed License: closed Type: model AI model by Hume https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GLM-4.6-2025-09-30 Tue, 30 Sep 2025 00:00:00 GMT Z.ai (Zhipu AI),Tsinghua University open Parameters: 357B License: open Type: model AI model by Z.ai (Zhipu AI),Tsinghua University https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Sora%202.0-2025-09-30 Tue, 30 Sep 2025 00:00:00 GMT OpenAI closed License: closed Type: model AI model by OpenAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Kandinsky%205.0%20Video%20Lite-2025-09-30 Tue, 30 Sep 2025 00:00:00 GMT Sber open Parameters: 2B License: open Type: model AI model by Sber https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Claude%20Sonnet%204.5-2025-09-29 Mon, 29 Sep 2025 00:00:00 GMT Anthropic closed License: closed Type: model AI model by Anthropic https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#NVIDIA%20Isaac%20GR00T%20N1.6-2025-09-29 Mon, 29 Sep 2025 00:00:00 GMT NVIDIA closed License: closed Type: model AI model by NVIDIA https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Cosmos-Transfer2.5-2B-2025-09-29 Mon, 29 Sep 2025 00:00:00 GMT NVIDIA open Parameters: 2B License: open Type: model AI model by NVIDIA https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Cosmos-Predict2.5-14B-2025-09-29 Mon, 29 Sep 2025 00:00:00 GMT NVIDIA closed Parameters: 14B License: closed Type: model AI model by NVIDIA https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Cosmos-Predict2.5%202B-2025-09-29 Mon, 29 Sep 2025 00:00:00 GMT NVIDIA open Parameters: 2B License: open Type: model AI model by NVIDIA https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#DeepSeek-V3.2-Exp-2025-09-29 Mon, 29 Sep 2025 00:00:00 GMT DeepSeek open Parameters: 671B License: open Type: model AI model by DeepSeek https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#MinerU2.5-2025-09-29 Mon, 29 Sep 2025 00:00:00 GMT Shanghai AI Lab,Peking University,Shanghai Jiao Tong University open Parameters: 1.2B License: open Type: model AI model by Shanghai AI Lab,Peking University,Shanghai Jiao Tong University https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Wan%202.5-2025-09-29 Mon, 29 Sep 2025 00:00:00 GMT Alibaba closed License: closed Type: model AI model by Alibaba https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Seedream%204.0-2025-09-28 Sun, 28 Sep 2025 00:00:00 GMT ByteDance closed License: closed Type: model AI model by ByteDance https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Kling%202.5%20Turbo-2025-09-26 Fri, 26 Sep 2025 00:00:00 GMT Kuaishou Technology closed License: closed Type: model AI model by Kuaishou Technology https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Suno%20v5-2025-09-25 Thu, 25 Sep 2025 00:00:00 GMT Suno closed License: closed Type: model AI model by Suno https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Gemini%20Robotics%201.5-2025-09-25 Thu, 25 Sep 2025 00:00:00 GMT Google DeepMind closed License: closed Type: model AI model by Google DeepMind https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Gemini%20Robotics-ER%201.5-2025-09-25 Thu, 25 Sep 2025 00:00:00 GMT Google DeepMind closed License: closed Type: model AI model by Google DeepMind https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GigaEmbeddings-2025-09-25 Thu, 25 Sep 2025 00:00:00 GMT Sber,Moscow Institute of Physics and Technology open Parameters: 3B License: open Type: model AI model by Sber,Moscow Institute of Physics and Technology https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#SimpleFold-2025-09-23 Tue, 23 Sep 2025 00:00:00 GMT Apple open Parameters: 3B License: open Type: model AI model by Apple https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#DeepSeek-V3.1-Terminus-2025-09-22 Mon, 22 Sep 2025 00:00:00 GMT DeepSeek open Parameters: 671B License: open Type: model AI model by DeepSeek https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Qwen3-Omni-Flash-2025-09-22 Mon, 22 Sep 2025 00:00:00 GMT Alibaba closed Parameters: 35.3B License: closed Type: model AI model by Alibaba https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Qwen3-Omni-30B-A3B-2025-09-22 Mon, 22 Sep 2025 00:00:00 GMT Alibaba open Parameters: 35.3B License: open Type: model AI model by Alibaba https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Grok%204%20Fast-2025-09-19 Fri, 19 Sep 2025 00:00:00 GMT xAI closed License: closed Type: model AI model by xAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Magistral%20Small%201.2-2025-09-18 Thu, 18 Sep 2025 00:00:00 GMT Mistral AI open Parameters: 24B License: open Type: model AI model by Mistral AI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Magistral%20Medium%201.2-2025-09-18 Thu, 18 Sep 2025 00:00:00 GMT Mistral AI closed License: closed Type: model AI model by Mistral AI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Granite-Docling-2025-09-17 Wed, 17 Sep 2025 00:00:00 GMT IBM open Parameters: 258M License: open Type: model AI model by IBM https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#AgentFounder-30B-2025-09-16 Tue, 16 Sep 2025 00:00:00 GMT Alibaba open Parameters: 30B License: open Type: model AI model by Alibaba https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Fabric%201.0-2025-09-15 Mon, 15 Sep 2025 00:00:00 GMT Veed closed License: closed Type: model AI model by Veed https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#GPT%E2%80%915-Codex-2025-09-15 Mon, 15 Sep 2025 00:00:00 GMT OpenAI closed License: closed Type: model AI model by OpenAI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Qwen3-Next-80B-A3B-2025-09-10 Wed, 10 Sep 2025 00:00:00 GMT Alibaba open Parameters: 80B License: open Type: model AI model by Alibaba https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Lucid%20Origin-2025-09-10 Wed, 10 Sep 2025 00:00:00 GMT Leonardo AI closed License: closed Type: model AI model by Leonardo AI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Ling-mini-base-2.0-20T-2025-09-10 Wed, 10 Sep 2025 00:00:00 GMT Ant Group open Parameters: 16B License: open Type: model AI model by Ant Group https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Ling-flash-base-2.0-20T-2025-09-10 Wed, 10 Sep 2025 00:00:00 GMT Ant Group open Parameters: 100B License: open Type: model AI model by Ant Group https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#K2%20Think-2025-09-09 Tue, 09 Sep 2025 00:00:00 GMT Mohamed bin Zayed University of Artificial Intelligence (MBZUAI),G42 open Parameters: 32B License: open Type: model AI model by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI),G42 https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Signal%20Processing%20Transformer-2025-09-09 Tue, 09 Sep 2025 00:00:00 GMT Softbank closed License: closed Type: model AI model by Softbank https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Qwen3-Max-2025-09-05 Fri, 05 Sep 2025 00:00:00 GMT Alibaba closed Parameters: 1T License: closed Type: model AI model by Alibaba https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#EmbeddingGemma-2025-09-05 Fri, 05 Sep 2025 00:00:00 GMT Google DeepMind open Parameters: 308M License: open Type: model AI model by Google DeepMind https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Chatterbox%20Multilingual-2025-09-04 Thu, 04 Sep 2025 00:00:00 GMT Resemble AI open License: open Type: model AI model by Resemble AI https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Apertus%208B-2025-09-02 Tue, 02 Sep 2025 00:00:00 GMT ETH Zurich,Ecole Polytechnique F´ed´erale de Lausanne (EPFL),Swiss National Supercomputing Centre (CSCS),Swisscom open Parameters: 8B License: open Type: model AI model by ETH Zurich,Ecole Polytechnique F´ed´erale de Lausanne (EPFL),Swiss National Supercomputing Centre (CSCS),Swisscom https://llm-timeline.duyet.net https://llm-timeline.duyet.net/#Apertus%2070B-2025-09-02 Tue, 02 Sep 2025 00:00:00 GMT ETH Zurich,Ecole Polytechnique F´ed´erale de Lausanne (EPFL),Swiss National Supercomputing Centre (CSCS),Swisscom open Parameters: 70B License: open Type: model AI model by ETH Zurich,Ecole Polytechnique F´ed´erale de Lausanne (EPFL),Swiss National Supercomputing Centre (CSCS),Swisscom