AI Briefing#model-release
The latest AI and developer news about #model-release, with the original source and a short summary.
The latest AI and developer news about #model-release, with the original source and a short summary.
Mica v0.1 4B is an Apache-2.0 licensed model that outputs calibrated probabilities for agent decisions, achieving 67.0 accuracy on held-out English sets while maintaining a small local footprint.
The model uses a single reinforcement-learning stage without prior domain-specific supervised fine-tuning.
The new Swift 1.5 27B model reduces thinking tokens by 58.5% while improving accuracy, with Flash Next offering a 1.8x speed increase.
Although its intelligence index is below average, it offers approximately 7x faster output speed and lower cost compared to models in the same price range.
Google has unveiled the Gemini 3.8 Flash TTS series, which enables voice generation and cloning via natural language prompts.
Next GenAi has released 'Rebekah 1.8B', an open-source LLM for edge deployment focused on data curation.
Alibaba Qwen has released Qwen-Image-2.1, an image generation model that reduces parameters to 7B and supports transparency generation.
Convai Innovations has open-sourced Laya, a non-autoregressive multilingual decision model that classifies states without text generation.
Cua has released CUA-S1, a small specialized model for computer-use decision-making, along with related open-source tools.
A test file referencing the next-generation model 'M3.1' was discovered in recent commits to MiniMax's official GitHub repository.
OpenRouter has officially released the multimodal model 'Unbiased Pareto' after completing stealth testing.
Wispr Advanced Interfaces Lab released 'Canto', a voice model that achieves low Word Error Rate (WER) in real-time transcription environments by applying GRPO reinforcement learning.
China Telecom has released Xing4.0-29B-A4B, a 29B MoE model trained on Ascend NPU.
The Agnes-3.0-Flash 33B model, released on Hugging Face, supports a hybrid attention architecture and a 262k context window.
DeepSeek released the 552B MoE model V4.1-Flash, replacing V4-Pro and optimizing KV cache to 890 bytes per token.
Sber released the MoE-based GigaChat-3.5-Reasoning, achieving DeepSeek V4 Flash-level performance while reducing inference token usage by 37%.
DeepSeek announced that it will officially launch v4.1 Flash on September 10, stating it outperforms v4 Pro across all metrics.
The OUI-1 model, fine-tuned from DiffusionGemma, achieved 71.7% on the Generative UI Benchmark, releasing high-quality UI generation technology capable of local execution.
A solo developer from India has released Arcle V1, an open-weight 5.8-billion-parameter multimodal model based on a single neural network.
IFM has open-sourced K2 Horizon, a suite of 6 models that exposes the entire process from pre-training to agent post-processing.
Ant Group has released Ling-3.0-flash-Fin, a 124B-parameter open-source model trained on financial data.
Google has released Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, models specialized for agentic workflows and cybersecurity.
Multiverse Computing has released Quasar 438B, an inference model that achieved the highest score among European models with 43 points on the Artificial Analysis Index.
The Qwen3-Coder-Next 4bit model, quantized using the calibration-free TQ method, has been released, reducing disk usage to 41.7GB.