open-source

Baidu open-sourced ERNIE 5.5 Flash Base on November 5. It is a Mixture-of-Experts base model pretrained on 33.5 trillion tokens, with 128K pretraining context and up to 1M tokens in fine-tuning and inference. Baidu claims it sets state-of-the-art results among models of the same size.

The IndiaAI Mission has released three indigenous foundation models: EKA-1 (26.3 billion parameters, trained on about 7.5 trillion tokens) from Soket AI Labs, and Dhīr 2B and Dhīr 5B from ZenteiQ AiTech Innovations. They were unveiled in New Delhi on November 3.

SenseTime open-sourced SenseNova-Loop-9B-Preview on October 30 under Apache 2.0. The multimodal model stores 32 decoder layers but runs 56 layer executions per forward pass, and SenseTime claims it matches models of around 27 billion parameters.

DeepSeek has published an engineering write-up and a GitHub prototype for Super-agent, a two-layer agent architecture built around a Router and parallel Workers. In a simulated environment of about 300 messages and 20 work requests, DeepSeek reports a median task completion time of 12.4 minutes and a 95th-percentile time of 15.7 minutes.

Cognition released the Devin Vulnerability Fixes plugin, made up of 3 MIT-licensed agent skills. They inventory, patch and test PHP, Node.js and self-managed GitLab deployments affected by this week's critical RCE flaws, then open 1 PR per repo.

A research team involving miHoYo's Honkai AI researchers and university collaborators has open-sourced GameScaling, a family of eight checkpoints from 800 million to 27 billion parameters. The models generate keyboard and mouse actions from visual observations and were built by continued pretraining of Qwen3.5 vision-language models.

MOSI Intelligence and OpenMOSS have open-sourced MOSS-Image, a 5B-parameter text-to-image model focused on dense text and JSON layout control. According to Pandaily, it quickly topped Hugging Face's text-to-image trending list.

Gradium has open-sourced Phonon, a family of on-device text-to-speech models. The English Phonon 31m has 31M parameters, ships under Apache 2.0, fits in 64.5 MB, and starts speaking in about 7 ms on a laptop CPU.

At its 1024 developer festival, iFLYTEK released Spark X2.6, a 293B-A30B mixture-of-experts model trained on domestic compute. The company also open-sourced the on-device Spark X2.6-4B and debuted the SparkAudio 1.0 speech model.

Xiaomi's SeerRay-Lab has open-sourced Xiaomi-Lens-0, a 4B multimodal model built on Qwen3.5-4B. The model emits mask tokens for pixel-level image regions in the middle of an answer. The lab also released a benchmark called LensBench.

IBM Research released Granite Timeseries Ensemble R1, a recipe that combines four pre-trained Granite forecasters into one median and quantile forecast with no retraining or fitted weights. On 97 GIFT-Eval configurations, the Probabilistic Ensemble scores 0.683 MASE, against 0.685 for the best single member.

NVIDIA has released Kumo Forecast (formerly NV-Tesseract Forecasting), a multivariate time-series forecasting model under Apache 2.0. It pairs a frozen MOMENT-1-large encoder of roughly 340 million parameters with a trainable forecasting head, optional nearest-neighbor retrieval over historical data (DARR), and a built-in interpretability framework.

Deepseek has open-sourced DeepInfer, the internal inference framework that serves its V4.1-Flash model, under the Apache License 2.0. The company says it will also open-source code for future models and has released deepseek-recipe to reproduce its API's token processing in self-hosted setups.

Hugging Face's Transformers v5.20.1 adds Beam, a model from Reflection AI that the release notes describe as token-efficient, frontier-scale, and aimed at coding and agentic work. The patch also fixes a regression that broke model loading on older torch versions. Beam's specifications and pricing were not disclosed in the release notes.

Ai2 has open-sourced AstaBrief 8B, a model fine-tuned from Qwen3-8B that turns a research question and retrieved literature excerpts into a cited report. It is live in Asta as Fast mode, which averages 51.1 seconds per report versus 178.5 seconds for the Claude-powered Thinking mode, according to Ai2. The weights and training data are public.

Xiaomi's MiMo team released MiMo-V2.6-Flash-RL, an efficiency-tier checkpoint in the MiMo-V2.6 series featuring a 309B-parameter (15B active) Mixture-of-Experts architecture, 1M-token context, and native support for text, image, video, and audio. The model uses a single mixed reinforcement learning run across coding, agentic, visual, and cybersecurity tasks rather than domain-specific training.

Xiaomi has released MiMo-V2.6-Flash, an open-source Mixture-of-Experts model with 309B total parameters and 15B activated per token, featuring a 1M-token context window and native multimodal capabilities. Priced at $0.14 per 1M input tokens and $0.28 per 1M output tokens, it targets agentic coding and long-horizon task workflows.

Yandex has released AliceAI-Foundation-80B-A3B-Base, an 80-billion-parameter hybrid MoE base model with 3 billion active parameters per token and a 262,144-token context window. The model was trained fully from scratch and, according to Yandex, outperforms larger open-source models on Russian-language factual and educational benchmarks.

Alibaba's Qwen team has open-sourced Qwen-Image-2.1, a 7B parameter unified model for text-to-image generation and image editing. The release adds native transparent (RGBA) image support and editing with up to 10 reference images.

Tencent has open-sourced AuK, a 1.5B-parameter foundation model for speech generation and editing that handles TTS, content editing, and audio enhancement through natural-language instructions. The release includes a distilled AuK-Flash variant for 4-step fast inference, both under MIT license.

DeepSeek has released DeepSeek-V4-Flash-Vision-Exp, its first experimental multimodal model in the V4 family, adding visual understanding to the V4-Flash architecture. The 305B-parameter model shows substantial gains on multimodal agent benchmarks while holding steady on text-only tasks.

Alibaba's Qwen team has released Qwen3.8-27B, a 27-billion-parameter dense vision-language model with 262,144-token native context extensible to 1 million tokens. The model shows gains over Qwen3.6-27B and Qwen3.7-Plus across coding, agentic, and multimodal benchmarks, according to Alibaba.

Meta CEO Mark Zuckerberg announced the company will open-source its Muse Spark 1.2 model and launch a new on-device model family called Muse Glimmer. The move positions Meta against closed-model rivals OpenAI and Anthropic and against Chinese open-weight labs like DeepSeek and Alibaba.

Microsoft Research released VibeVoice-ASR-BitNet, a quantized 1.58GB version of its VibeVoice-ASR speech recognition model that achieves real-time inference (RTF < 1) on as few as 3 CPU threads. The model runs 1.6-2.3x faster than Whisper.cpp on commodity x86 and ARM hardware, with a modest accuracy tradeoff.

Within days, Moonshot AI and Alibaba unveiled what they claim are frontier-class models. Moonshot's Kimi K3, at 2.8 trillion parameters, and Alibaba's Qwen3.8, at 2.4 trillion parameters, will both be released as open-weight models with full weights available for download.

LM Studio released Bionic, a Mac app that runs open-source AI models locally or via cloud for coding, document processing, and research tasks. The app includes offline voice transcription using Mistral's Voxtral model and supports models like GLM 5.2 and Kimi K2.7 Code for codebase editing.

Cohere and Cohere Labs released Cohere Transcribe Arabic, a 2B parameter automatic speech recognition model optimized for Arabic dialects and Arabic-English code-switching. The open-source model achieves a 25.9% average word error rate across major Arabic ASR benchmarks, outperforming models up to 30B parameters.

Chinese AI models including DeepSeek and Z.ai have captured over 30% of weekly token usage by U.S. companies on OpenRouter since February 2025, up from 4.5% in the first half of the year. The shift comes as companies seek alternatives 60-90% cheaper than leading models from OpenAI and Anthropic, while Chinese models close the performance gap to within 6-9 months of U.S. frontier systems.

Nex AGI has released Nex-N2-Mini, an open-source agentic mixture-of-experts model with a 262K-token context window. The model accepts text and image inputs and is priced at $0.025 per 1M input tokens and $0.10 per 1M output tokens.

Mistral AI has released Leanstral 1.5, an open-source 119B parameter mixture-of-experts model designed specifically for Lean 4 proof assistance. The model features 128 experts with 4 active per token (6.5B activated parameters), a 256k token context window, and multimodal input capabilities.

Portugal has released Amália, its first national AI model trained specifically for European Portuguese. Built on EuroLLM-9B with 9 billion parameters, the model is fully open-source with weights, datasets, and code published under an open license. The government has committed €5.5m in initial funding through 2027.

DeepReinforce has released Ornith-1.0, an MIT-licensed model designed for agentic coding tasks with variants ranging from 9B to 397B parameters. Built on top of Apache 2.0-licensed Gemma 4 and Qwen 3.5 base models, the company claims it achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks.

DeepSeek has released two new MoE models: DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated). Both models support a one million token context window and use a hybrid attention architecture that requires only 27% of single-token inference FLOPs and 10% of KV cache compared to DeepSeek-V3.2.

The open model ecosystem is diversifying beyond its previous Chinese dominance, with companies like NVIDIA, Cohere, Poolside, and Zyphra releasing models under permissive licenses. NVIDIA's 550B parameter Nemotron-3-Ultra uses LatentMoE architecture and switches to the OpenMDW license, while Cohere released Command A+ as a 218B-A25B MoE under Apache 2.0.

DeepSeek has released two Mixture-of-Experts models: V4-Pro with 1.6 trillion parameters (49B activated) and V4-Flash with 284B parameters (13B activated), both supporting 1 million token context windows. V4-Pro requires only 27% of inference FLOPs and 10% of KV cache compared to V3.2 at 1M token context, trained on over 32 trillion tokens.

Zhipu AI's open-source GLM-5.2 model scores within one percentage point of Anthropic's Claude Opus 4.8 on a key agentic benchmark while costing approximately one-fifth as much. The release comes as U.S. government restrictions limit access to Anthropic's Fable and OpenAI's GPT-5.6 models.

Z.ai's GLM-5.2 performs on par with Claude Opus 4.8 and OpenAI's GPT-5.5 in cybersecurity benchmarks while costing roughly half as much to run. Security evaluations from Graphistry and Semgrep confirm the open-weight model's capabilities in vulnerability discovery and cyber investigation, raising concerns about accessibility of advanced hacking tools.

Trail of Bits launched Patch the Planet, a security initiative using OpenAI's GPT-5.5-Cyber model to find and fix bugs in critical open-source projects. The first week produced 64 pull requests and 51 issues across 19 projects including cURL, Python, PyPI, and Sigstore, with 37 patches already merged.

Baidu has released Unlimited-OCR, a 3 billion parameter model for optical character recognition and document parsing. The model supports single-page and multi-page document processing with a 32,768 token context window and runs on NVIDIA GPUs using bfloat16 precision.

Poolside has released Laguna M.1, a 225B total parameter Mixture-of-Experts model with 23B activated parameters per token, designed for agentic coding tasks. The model scores 74.6% on SWE-bench Verified and 63.1% on SWE-bench Multilingual, released under Apache 2.0 license.

Mistral AI has released Leanstral, a sparse 120B model with 6B active parameters designed specifically for the Lean 4 proof assistant. The model is available under Apache 2.0 license with free API access and achieves a 26.3 FLTEval score at pass@2, outperforming Claude Sonnet 4.6 while costing $36 versus $549.

Mistral has released Mistral 3, including Mistral Large 3—a sparse mixture-of-experts model with 41B active and 675B total parameters—and three Ministral 3 edge models (3B, 8B, 14B). All models are released under Apache 2.0 license with multimodal capabilities and are available today on multiple platforms.

Zhipu AI released GLM-5.2, a 753 billion parameter model with a 1 million token context window. The model scores 62.1% on SWE-bench Pro and introduces IndexShare architecture that reduces per-token FLOPs by 2.9× at 1M context length. Released under MIT license with no regional restrictions.

Z.AI has released GLM-5.2, an open-source model with a 1M-token context window under an MIT license. On FrontierSWE, a long-horizon coding benchmark, GLM-5.2 trails Claude Opus 4.8 by 1% while outperforming GPT-5.5 by 1%, and achieves 81.0 on Terminal-Bench 2.1 compared to Opus 4.8's 85.0.

Zhipu AI has released GLM-5.2, a 753 billion parameter model featuring a 1 million token context window and MIT open-source license. The model scores 62.1% on SWE-bench Pro and 91.2% on GPQA-Diamond, with flexible reasoning effort levels for coding tasks.

Microsoft released FastContext-1.0, a lightweight repository-exploration subagent for LLM coding agents spanning 4B to 30B parameters. The model reduced main-agent token consumption by up to 60% while improving end-to-end resolution rates by up to 5.5% on SWE-bench Pro when integrated with agents like GPT-5.4 and GLM-5.1.

Google DeepMind has released Gemma 4 12B, a 12-billion parameter multimodal model that runs locally on laptops with 16GB of RAM. The model eliminates separate vision and audio encoders, processing raw inputs directly through its language model backbone under an Apache 2.0 license.

Amazon has released an open-source testing framework for Nova Sonic voice agents that automates multi-turn conversation evaluation without requiring human testers. The harness uses LLM-as-judge techniques to assess voice agents across six metrics including goal achievement, response accuracy, and tool usage, addressing a critical QA bottleneck in voice AI development.

Google DeepMind has released Gemma 4 12B Unified, an encoder-free multimodal model that processes text, images, and audio through a single decoder-only transformer. The model features 11.95 billion parameters, a 256K token context window, and achieves 77.2% on MMLU Pro and 72.0% on LiveCodeBench v6.

ByteDance released Bernini-R, an open-source video generation and editing model that combines an MLLM-based semantic planner with a DiT-based renderer. The model requires Hopper-class GPUs (H100/H800/H200) for optimal performance and supports multiple tasks including text-to-video, video editing, and reference-guided generation.