NAll news

All news

Thursday37 items

  1. QbitAIModel releases
    AI score66

    China Telecom Xingchen Releases and Open-Sources Intent Understanding Model Xing4.0-29B-Intent

    The Xingchen General Artificial Intelligence Laboratory of China Telecom Artificial Intelligence Technology Co., Ltd. has released and open-sourced the dedicated intent understanding model Xing4.0-29B-Intent, simultaneously opening 2.753 million training corpus entries, model weights, and inference images. The model is fine-tuned based on Xing4.0-29B-A4B and adopts a unified schema-driven semantic parsing framework, supporting both intent recognition and parameter extraction. It outputs field-constrained JSON, with an average output of a few dozen tokens, and the full result response latency can reach the hundreds-of-milliseconds level. The model has been applied in conversational intent recognition and intelligent ticket routing for Telecom's ten-thousand-line customer service, as well as spoken command parsing for smart home speakers. The code is open under the Apache 2.0 license, and the dataset and weights will be released simultaneously on Hugging Face.

    #Model releases#Open-source ecosystem#Reasoning

  2. MarkTechPostModel releases
    AI score62

    IBM Releases Granite Speech 5.0 2B Multilingual Speech Transcription and Translation Model

    IBM has released the open-source speech language model Granite Speech 5.0 2B under the Apache 2.0 license. It supports speech-to-text in 6 languages and also offers speech translation capabilities. On the Open ASR Leaderboard, the model ranks second among open-weight models with an average English WER of 5.68%, behind only Qwen3-ASR-1.7B at 5.40%. Compared with 4.1 2B, it adds language ID prompting, keyword biasing of up to 2000 words, and on-demand punctuation and capitalization, and can run on Transformers, vLLM, llama.cpp, and mlx-audio.

    #Model releases#Speech#Open-source ecosystem

  3. English text is being prepared.Model releases
    AI score68

    iFlytek Spark X2.6 cloud-device integrated large model released, achieving breakthroughs in coding and agent capabilities based on fully domestic computing power

    iFlytek released the iFlytek Spark X2.6 cloud-device integrated large model at the 1024 Developer Festival, claiming breakthroughs in coding and agent capabilities based on fully domestic computing power for the first time. The model adopts a 293B-A30B parameter MoE architecture, expands context length to 1M, and supports custom security policies by task scenario. The simultaneously released Spark X2.6-4B open-source on-device large model improves coding, agent, long-text and multilingual capabilities, adds three tiers of reasoning modes, and offers free reasoning for a limited time; the X2.6 series is now available for trial on the iFlytek Spark official website.

    Same story · 7 reports

    #Model releases#Agents#English text is being prepared.

  4. IT Home · Artificial IntelligenceModel releases
    AI score63

    iFlytek Releases SparkAudio 1.0, a Spark Speech Foundation Model

    At the 1024 Developer Festival, iFlytek announced the debut of SparkAudio 1.0, a Spark speech foundation model that integrates capabilities such as machine translation, speech recognition, and semantic understanding, and claims outstanding performance in real-time speaker separation, environmental understanding, audio understanding, and emotion recognition. The model has been fully launched in iFlytek Input Method 16.0, iFlytek Hearing, iFlytek Office Notebook, and iFlytek Translator, supporting arbitrary switching between Chinese and English and real-time speaker recognition. iFlytek also released the iFlytek AI Recording Ring, equipped with a 4 + 1 ring microphone array and Wi-Fi 6 transmission, claiming a comprehensive recognition accuracy rate of over 96% and that transcribing 2 hours of recording takes only 2 minutes.

    Same story · 2 reports

    #Model releases#Speech#Multimodal

  5. indigoModel releases
    AI score67

    Mirendil Releases Mirendil-Base Code Foundation Model, Pretraining Completed in 3.5 Months

    Mirendil has released its code foundation model Mirendil-Base, pretrained from scratch by one researcher leading a group of Agents over 3.5 months at a cost of under $10 million, producing a model with 100–500 billion parameters. It used only public data—no distillation, no synthetic data, and no purchased data—and more than 99% of the code was written by Agents. On code prediction loss it outperforms GLM-5.3, DeepSeek-V4-Pro-Base, and Nemotron-550B, but falls short of Kimi-K3. The model does not release its weights, has no API, and has no task evaluations such as SWE-bench yet; the author views this as more of an efficiency win for a senior researcher plus Agents, and the method relies on third-party Coding Agents and is easy to replicate.

    Same story · 2 reports

    #Model releases#Agents#English text is being prepared.

  6. Phoenix TechnologyModel releases
    AI score62

    ElevenLabs Releases Eleven Realtime v1 Mini and Max Real-Time Speech Models

    On October 20, ElevenLabs released the Eleven Realtime series of real-time speech models, including v1 Mini for natural conversation and v1 Max for customer service and multi-step tasks, both supporting more than 90 languages and 17,000 voices. According to Artificial Analysis testing, Max achieved 98.8% accuracy on the Big Bench Audio speech reasoning benchmark, ranking third, with an average time to first audio response of 0.57 seconds, while Mini achieved 75.9% accuracy and 0.62-second response time. Both models can be used through the ElevenAgents platform or the Realtime API compatible with the OpenAI Realtime specification, and currently only support WebSocket connections; in the author's hands-on test, when building an assistant with Architect, the Chinese accent was noticeable, responses tended to be long, and latency was relatively high.

    Same story · 7 reports

    #Model releases#Speech#Agents

  7. The ElecModel releases
    AI score57

    Kakao Open-Sources Kanana-2-vision-encoder Vision Encoders in 426 Million and 98 Million Parameter Versions

    Kakao has open-sourced two versions of its self-developed image recognition model Kanana-2-vision-encoder: a medium version with 426 million parameters and a tiny version with 98 million parameters, both now available on Hugging Face. Kakao says that across 42 benchmarks spanning nine major categories, including general visual question answering, English-Korean OCR, document understanding and object grounding, the two models achieved the highest performance compared with Google SigLIP 2, Meta Perception Encoder, Apple AIMv2 and Alibaba Qwen3.5-ViT. The tiny version outperforms competitors with more than 400 million parameters on some benchmarks, with particularly notable gaps in Korean recognition, document understanding and object grounding. During training, the use of the self-developed Kanana-2 language model improved training efficiency by about 4x, and architectural improvements for high-resolution image processing reduced latency by more than 60%; the models are released under the Kanana Open License, permitting research, development and commercial use.

    #Model releases#Open-source ecosystem#Multimodal

  8. Testing CatalogModel releases
    AI score68

    Recursive Releases Flint, Its First Open-Weight Coding Model

    Recursive has released Flint, its first open-weight coding model, under the Apache 2.0 license. It is a 40B-parameter MoE model aimed at local coding scenarios. Recursive says it reaches over 220 tokens per second on a 64GB RTX Spark, 2x faster than comparable models. Its automated research system, Eureka Machine, also generated custom GPU code that sped up Flint on the RTX Spark by 55% without changing the weights.

    Quoted postRichard Socher · @RichardSocher

    Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

    Same story · 9 reports

    #Model releases#Open-source ecosystem#English text is being prepared.

  9. Dongseok Ma NLPModel releases
    AI score58

    Base Labs Releases mammoth-1 and elephant-1 Decision Models and Open-Sources Weights

    Base Labs has released two decision models, mammoth-1 and elephant-1, stating that mammoth-1 is the world's top-ranked decision model and elephant-1 is the best model under 15B and surpasses Jev. Base Labs has open-sourced all model weights and provided a technical report detailing its training and data methods.

    Quoted postBase Labs · @baselabs

    Today we release mammoth-1 and elephant-1! We post-trained mammoth-1, the #1 ranked best decision model in the world, and elephant-1, which is the best model under 15B (and beats Jev). We're open-sourcing all model weights and provide a technical report detailing all our training and data methods.

    #Model releases#Open-source ecosystem

  10. Shao MengModel releases
    AI score75

    Recursive releases its first model Flint: local coding MoE, 125+ tokens/s on a MacBook

    Recursive releases its first model Flint, a MoE with 40 billion total parameters and 2.5 billion activated parameters per token, aimed at local Agent coding, with weights open-sourced under Apache 2.0. With short context, it exceeds 125 tokens per second on an M4 Pro MacBook Pro and over 150 on a 64GB RTX Spark, and supports llama.cpp, Ollama, LM Studio, vLLM, SGLang, and Unsloth. Nearly every stage of its R&D involved the automated research system Eureka Machine, including architecture search, quantizing the output layer to 4 bit with offline distillation, and fusing Metal kernels for M4/M5.

    Quoted postRichard Socher · @RichardSocher

    Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

    Same story · 9 reports

    #Model releases#Open-source ecosystem#Agents

  11. OpenRouterFirst-partyModel releases
    AI score62

    Recursive's first open-source model Flint launches on OpenRouter, a 40B MoE built specifically for coding agents

    Recursive's first model Flint has launched on OpenRouter. It is an open-weight 40B MoE model (about 2.5B active parameters), built by the company's automated research system Eureka Machine, aimed at coding agents, and supports 262K context, reasoning, and tool calling. Recursive reports that it scores 54.9% on Terminal-Bench 2.1, 89.7% on SWE-bench Pro v2, and 74.2% on SWE-bench Verified, and is free to use during its launch period.

    Quoted postRichard Socher · @RichardSocher

    Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

    #Product updates#Open-source ecosystem#Agents

  12. Rohan PaulModel releases
    AI score76

    Recursive releases Flint, a 40B MoE coding model, at 125+ tokens/s on a MacBook Pro

    Recursive has released its first model, Flint, a 40B mixture-of-experts coding model under the Apache 2.0 license, with weights now open on Hugging Face, built from scratch in under 4 months by the automated research system Eureka Machine. Flint generates code at 125+ tokens/s on a MacBook Pro and reaches 150+ tokens/s on an RTX Spark. On a 48GB MacBook, its average quality is about 81%, compared with about 85% for Qwen3.8 27B, but in 128K-token long sessions it is about 7 times faster than the latter.

    Quoted postRichard Socher · @RichardSocher

    Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

    Same story · 9 reports

    #Model releases#English text is being prepared.#English text is being prepared.

  13. Ayi AI NotesModel releases
    AI score67

    MemT-1.0, an open-source native-memory large model, cuts 256K memory usage to a constant 40M

    MemT-1.0 is open-sourced, connecting a native memory module in parallel within the Transformer to turn memory into a state in the forward computation flow, releasing up to 288B weights and a 66-page technical report. The authors claim that under a 256K ultra-long history, memory overhead drops from 5.06G to a constant 40M, KV cache is reduced by more than 120 times, and first-token latency stays stable at just over 100 milliseconds. The team also wrote more than a thousand code-correction experiences into the internal memory state, which do not occupy prompt tokens, and the pass rate on a standard code test set increased by 8% for three consecutive runs.

    Same story · 3 reports

    #Model releases#Open-source ecosystem#Agents

  14. MarkTechPostModel releases
    AI score70

    Recursive Releases Open-Weight Coding Model Flint, a 40B MoE with Only 2.5B Active Parameters

    Recursive Superintelligence has released its first model, Flint, a 40B MoE model aimed at coding agents with roughly 2.5B active parameters per token; the weights are released under the Apache 2.0 license on Hugging Face. The model card lists a context of 262,144 tokens, with 4-bit weights at about 23 GB, runnable on a 48 GB Apple Silicon Mac or a 48 GB NVIDIA GPU. The company self-reports SWE-bench Verified at 74.2% and IFEval at 95.1%, while Terminal-Bench 2.1 stands at 54.9%, below dense Qwen 3.8's 73.0%. Recursive says its automated research system Eureka Machine contributed to the design of the architecture, quantization, and kernels.

    Same story · 9 reports

    #Model releases#Open-source ecosystem#Agents

  15. TechCrunch AIModel releases
    AI score69

    Recursive releases laptop-class coding model Flint, says it was built mainly with AI involvement

    AI startup Recursive has released its first model, Flint, a coding model that runs offline on a laptop and is freely downloadable on Hugging Face under the Apache 2.0 license; the company says its development process was mainly carried out with the involvement of its in-house Eureka Machine. Flint is an MoE model that activates only 2.5B of its 40B parameters at a time, generates at about 131 tokens/sec on a MacBook Pro with an M4 Pro, and outperforms other models under 50B parameters on SWE-Bench Verified. Recursive says Flint requires more guidance than large paid coding assistants and recommends breaking large tasks into small steps.

    Same story · 9 reports

    #Model releases#English text is being prepared.#English text is being prepared.

  16. Hugging FaceFirst-partyModel releases
    AI score73

    NVIDIA Releases Kumo Relational Relational Foundation Model, Predicts Without Training

    NVIDIA has released Kumo Relational, a pretrained relational foundation model that can make predictions directly on enterprise databases in a single forward pass, without task-specific training, feature engineering, or table flattening. Across 41 tasks on 15 databases—RelBench V1/V2, SAP SALT, and 4DBInfer—it outperforms supervised relational deep learning models and gradient-boosted tree pipelines built with expert feature engineering, achieving an 8% higher MRR than the AutoGluon ensemble on SAP SALT. The model is available through NVIDIA's open-source GPU-native structured-data-models library or NVIDIA NIM, and supports databases with 500B+ rows.

    #Model releases#English text is being prepared.#Agents

  17. ZhidongxiModel releases
    AI score70

    NVIDIA Open-Sources Surgical World Model Cosmos3-H-Dreams, Enabling Real-Time Surgical Scene Generation on a Single GPU

    NVIDIA has released and open-sourced the real-time generative surgical world model Cosmos3-H-Dreams, which can generate surgical scenes in real time based on robot operation instructions, simulating instrument movement, tissue retraction, and deformation. The model is built on the 16-billion-parameter Cosmos3-H-Embodied-Nano, with training data comprising approximately 526 minutes of real surgical video selected from 4,792 surgical recordings, and it supports 28-dimensional action input. Using teacher-student distillation and the FlashDreams streaming inference framework, the model requires only a single NVIDIA RTX PRO 6000 GPU for real-time interaction, and two student model versions with 4-step and 2-step denoising have been open-sourced.

    Same story · 4 reports

    #Model releases#English text is being prepared.#Embodied AI

  18. Baseten Base Labs: Model ResearchFirst-partyModel releases
    AI score68

    Baseten Base Labs releases decision models mammoth-1 and elephant-1

    Baseten's Base Labs releases two decision models, elephant-1 (9B) and mammoth-1 (27B), both fully fine-tuned from Qwen3.5, with model weights publicly released. mammoth-1 ranks first among 208 models on the Jev Decision Index 0.4 with a score of 71.24, leading the second-place Perplexity Decider v1.2 by 1.8 points; elephant-1 scores 62.65 on that leaderboard and ties for first with d1-ultra on the Artificial Analysis Decision Index with a score of 65. Training used the eighth version of the data mixture with 6.21M samples, and replaced the full-attention layers with a block bidirectional attention mask; under fixed data, this change raised the public 10k-sample score from 44.8 to 48.6.

    Max Kirkby

    Same story · 2 reports

    #Model releases#Reasoning#English text is being prepared.

  20. kling_aiFirst-partyModel releases
    AI score60

    English translation is being prepared.

    Quoted postMorphic · @morphic

    Introducing Kling 4.0 on Morphic. Get 10% discount on Kling 4.0 generations through November 20 on Morphic. Native 30-second generations, up to 10 keyframes, 15 references, consistent on-screen text, multilingual support. 4K and 10-bit HDR. https://t.co/KEJsRjqn5m https://t.co/ltkDtrfGCv

    #Model releases#Video#Multimodal