#Model release

AIModelScope has released NeoHorse-Jev-4B, a compact open model that converts application states into structured decisions and probabilities. It scores 77.70 across six text decision benchmark groups, ranking first among four open-weight models with complete results in the comparison. Its prefill-only inference supports Choice, Noul, and Score primitives, accepts text or a single image with text, and is available under Apache 2.0 for deployment via vLLM, SGLang, Python, CLI, or HTTP.

AIOdyssey introduced Agora-2, its next-generation multi-agent world model, which supports up to 20 humans and agents interacting in a shared environment simulated in real time. A multiplayer research preview is available to try now.

AIOdyssey's research preview of Agora-2 lets up to 4 humans join forces against 16 agents in an environment simulated entirely in real time by a world model. The post states there is no game engine underneath.

AIGoogle's Philipp Schmid announced Gemini 3.8 Flash, recommending Gemini for multimodal understanding. A quoted post by Spencer Schiff reported that frontier models struggled to match correct names to people in a drawing, offering a visual test for future models.

AIModelScope released Qwen-Image-2.1-Fun-Controlnet-Union, a single checkpoint adding eight structural controls, including Canny, Depth, Pose, and Scribble, plus inpainting to Qwen-Image 2.1. Control and inpainting share one branch with 16 injection points across every second Transformer block, keeping the base model frozen and requiring no checkpoint switching. It runs at guidance scale 1.0 with CFG-distilled sampling and prefix KV caching, and is available under the Qwen Research License with base Qwen-Image 2.1 weights required.

AIMeta has published a recap of its Meta Connect event, highlighting the latest Muse updates, new AI glasses, and its Meta VR Glasses. The post itself gives no specifics on features, specifications, or pricing.

AIAnt Ling's Ming-Image-0.1-Design, a text-to-image model built for graphic-design output and legible text rendering, is now available via Novita on OpenRouter. Novita launched it as a Day-0 partner with a free 14-day offer.

AIAs of Sep 23 at 9:00 AM PT, Ling-3.0-flash-VL on OpenRouter switched from free to paid access. Limited-time discounted prices are $0.021 per 1M input tokens, $0.063 per 1M output tokens, and $0.0042 per 1M cache-hit tokens.

AIBlack Forest Labs says FLUX 3 Action can control a simulated drone to carry out a specific requested action. The post offers no further details on benchmarks, availability, or implementation.

AIBlack Forest Labs says FLUX 3 Action takes recent camera frames, the system's current state, and a task description to return the next 32 actions. The model also predicts how the scene will change, and runs a continuous observe-plan-act-adjust loop. It can recover from its own mistakes, according to the company.

AIBlack Forest Labs says FLUX 3 Action builds on the same image, video, and audio pretraining as FLUX 3 but uses a smaller architecture. The company attributes this smaller size to more efficient representations learned through its Self-Flow research. During midtraining, the model was trained to predict actions and future frames together.

AIBlack Forest Labs says its FLUX 3 Action, a single-step 7B checkpoint, outperforms every other open policy on RoboLab. It processes each second of robot motion 1.45× to 1.66× faster than Pi0.5, and uses a 2.13-second action horizon versus Pi0.5's 1 second. The company adds that its guidance-distilled checkpoint raises the state-of-the-art RoboLab success rate while running 2.85× to 3.15× faster than the previous leading open WAM.

AIBlack Forest Labs says FLUX 3 Action is an open-weights 7B world action model that ranks first on the RoboLab benchmark. The company says it outperforms the previous best open model by 6.1 percentage points while using 56% fewer parameters and running up to 3.95x faster. The model predicts video and actions together, and the company is releasing the weights, code, fine-tuning recipe, benchmarks, and examples. It also integrated the model into Hugging Face's LeRobot with NVIDIA, with edge deployment on NVIDIA Jetson.

Why it matters: The release pairs benchmark results with the trade-off it claims to remove between world action model performance and VLA speed, which is useful context for robotics teams weighing open models.

AIGoogle AI launched Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, which it describes as its most expressive audio models yet. The models support custom voices across 100+ languages, 2,000+ prebuilt voices, multi-speaker conversations, line-by-line delivery control, and cues such as <laughs> and |mhm|. Google positions Flash TTS for bespoke voices in gaming, audiobooks, and podcasts, and Flash-Lite TTS for near real-time voice agents, high-volume dubbing, and bulk audio.

AIGoogle DeepMind has introduced two text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS. Gemini 3.8 Flash TTS lets users design unique voices with distinct accents and characteristics, while Flash-Lite TTS is built for efficiency and scale, offering user-created styles or a production-ready voice library.

AIGoogle introduced Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, describing them as its most expressive audio generation models yet. The models are designed to help creators, developers, and enterprises build richer, more expressive audio experiences, and can be tried through the Gemini API and AI Studio.

AINVIDIA's Nemotron 3 Diarization is now available on ModelScope, labeling speakers and timestamps in streaming audio for up to eight speaker slots per conversation. The 99.2M-parameter model uses an end-to-end streaming architecture built on NVIDIA's Streaming Sortformer, running on Ampere, Hopper, and Blackwell GPUs via NeMo Speech C++. It is designed to pair with existing ASR systems such as Nemotron ASR, Parakeet, Canary, or Whisper to produce speaker-attributed transcripts.

AITeleOCR, a lightweight 1.2B vision-language model released under Apache 2.0, parses digital PDFs and warped phone photos without a separate dewarping model. It scores 96.87 overall on OmniDocBench v1.6, the highest among listed specialized VLMs, and ranks #1 in the ICDAR 2026 Sci-ImageMiner Challenge. It supports structured parsing of text, tables, formulas, layouts, and reading order, with synchronous or asynchronous vLLM inference.

AIAlibaba's Qwen team released Qwen-Audio-3.1, upgrading its ASR, TTS and Realtime models and adding TTS-Next and ASR-Next. The post says TTS prices fell about 70%, Realtime about 85%, and ASR up to 95%. More APIs are coming soon.

AIXiaomi has released MiMo-V2.6 as an open model family under the MIT License, designed for large-scale reinforcement learning. MiMo-V2.6-Pro scores 46 on the Artificial Analysis Intelligence Index, with 71.9 on DeepSWE v1.1, 89.9 on Terminal-Bench 2.1, and 82.0 on OSWorld-Verified. The 1.02T-parameter MoE activates 42B parameters and supports text, image, video, and audio input with a 1M-token context.

AIswyx reports that Claude Opus 5.5 is now the default model for Latent Space's AINews, after side-by-side testing against Sol showed more concise and tasteful reporting with less "slopese" than Opus 5. The post links to the AINews issue titled "Claude Opus 5.5: The New Default."

AIShanghai AI Lab and SJTU's LUMIA Lab released NCP-ArchPreview, an 8.9B open-weight language model under Apache 2.0. The model reportedly reaches OLMo-3-7B's final Stage 1 loss using 51.3% of the tokens from the 5.73T Dolma 3 corpus, a 1.95× convergence gain. Its concept module jointly predicts tokens and concepts, and domain adaptation updates only its 17M parameters while the token backbone stays frozen.