open-weights

NVIDIA has published NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4, a 30-billion-parameter Mixture-of-Experts model with only 3B active parameters, a hybrid Mamba-2/MoE/Attention architecture, and support for up to 1 million tokens of context. The NVFP4-quantized checkpoint is designed to run on a single DGX Spark (GB10) or H100 GPU.

Alibaba announced Qwen3.8-Max, a 2.4T-parameter flagship model targeting coding and long-horizon agentic work, with open weights promised for next week alongside Qwen3.8-27B. The model posted strong third-party benchmark results, ranking #4 in Frontend Code Arena and matching Claude Opus 4.7 on the Vals Index at roughly 2.3x lower cost.

Thinking Machines has released Inkling Small, an open-weights reasoning model with 276 billion total parameters but only 12 billion active. According to Artificial Analysis, it scores nearly as high as the company's larger Inkling model while using roughly a third of the parameters and far fewer output tokens per task.

Moonshot AI has released Kimi K3, a 2.8 trillion parameter model with 1 million token context window and native multimodal input. The model ranks #1 in Frontend Code Arena and #9 in Text Arena, with pricing at $3 per million input tokens and $15 per million output tokens—comparable to Claude Sonnet 5 pricing while delivering performance the company claims is near Claude Opus 4.8 and GPT-5.5.

Thinking Machines Lab released Inkling, a mixture-of-experts model with 975B total parameters and 41B active parameters, trained on 45 trillion tokens across text, images, audio, and video. The Apache 2.0-licensed model supports up to 1M context and debuts alongside Inkling-Small (276B-A12B), marking what observers call the strongest US-based open-weight release to date.

Moonshot AI released Kimi K3 on July 16, 2026, a 2.8 trillion parameter mixture-of-experts model that ranks #2 on the Vals AI index and #3 on Artificial Analysis's Intelligence Index. The company will release the model's weights on July 27, making it the strongest open-weight model to date, surpassing all previous open releases including DeepSeek R1.

Alibaba unveiled Qwen3.8 at the World Artificial Intelligence Conference in Shanghai, claiming the 2.4 trillion parameter model ranks second only to Anthropic's Fable 5. The company provided no benchmark scores, model card, or independent verification to support the claim.

Black Forest Labs has released FLUX.2, a family of image generation models including a 32B parameter open-weight variant. The models support editing at up to 4 megapixel resolution and can reference up to 10 images simultaneously for character and style consistency.

Black Forest Labs released FLUX.1 Krea [dev], an open-weights text-to-image model developed with Krea AI. The model was specifically trained to generate realistic images that avoid the oversaturated textures common in AI-generated imagery, achieving performance on par with FLUX1.1 [pro] in human preference tests.

Chinese AI lab Moonshot AI released Kimi K3, a 2.8 trillion parameter model priced at $3 per million input tokens and $15 per million output tokens. The model is currently available via API, with open weights promised by July 27, 2026. This represents the most expensive pricing from a Chinese AI lab to date, matching Anthropic's Claude Sonnet series.

NVIDIA's Nemotron-3-Embed-8B-BF16 ranks #1 on the RTEB leaderboard with a 78.5% score, while the 1B variant reduces error rate by 27% over its predecessor. The open-weight models feature 32k context windows and production-ready deployment options including a Blackwell-optimized NVFP4 variant.

Thinking Machines Lab released Inkling, a Mixture-of-Experts transformer with 975B total parameters and 41B active parameters, trained on 45 trillion tokens of text, images, audio and video. The Apache-2.0 licensed model is designed as a base for fine-tuning rather than a frontier model.

Krea.ai released Krea 2 Turbo, a 12-billion parameter diffusion transformer model for text-to-image generation. The open-weight model generates images in 8 inference steps and supports resolutions up to 2048x2048 pixels.

Z.ai released GLM-5.2 on June 16, 2026, the first open-weight model to match proprietary models like Claude Opus 4.8 on agent benchmarks. The MIT-licensed model closes the performance gap to 6.8 months behind frontier labs, down from expected 9+ months as compute scales.

Mistral AI has released Voxtral TTS, a 4B parameter text-to-speech model supporting 9 languages including English, French, German, Spanish, Dutch, Portuguese, Italian, Hindi, and Arabic. The model achieves 70ms latency for typical inputs and can clone voices from as little as 3 seconds of audio, priced at $0.016 per 1,000 characters.

Google has released DiffusionGemma 26B, an open-weight text generation model under Apache 2 license. The model generates over 500 tokens/second according to testing on NVIDIA's free NIM API, where it produced 2,409 tokens in 4.4 seconds.

NVIDIA released Nemotron-3-Ultra-550B-A55B-NVFP4, a 550B parameter model with 55B active parameters, featuring a 1M token context window and configurable reasoning mode. The model uses a hybrid LatentMoE architecture combining Mamba-2, Mixture-of-Experts, and Attention layers with Multi-Token Prediction, trained with NVIDIA's NVFP4 quantization-aware approach.

NVIDIA released Nemotron-3-Ultra-550B, a frontier-scale model with 550B total parameters (55B active) and up to 1M token context window. The model uses a hybrid LatentMoE architecture combining Mamba-2, MoE, and attention layers with Multi-Token Prediction, trained with NVFP4 quantization-aware methods from December 2025 to April 2026.

NVIDIA released Nemotron 3.5 ASR, a 600M-parameter speech-to-text model supporting 40 language-locales from a single checkpoint. The model achieves 0.07 seconds to final transcript after speech ends and ranks 2nd in latency among streaming ASR models according to Artificial Analysis benchmarks.

Google DeepMind released Gemma 4, a family of open-weight multimodal models ranging from 2.3B to 30.7B parameters. The flagship 12B Unified model eliminates separate encoders, processing text, images, audio, and video directly through a single decoder-only transformer with up to 256K token context window.

Stability AI has released Stable Audio 3.0, a model family for audio generation trained on fully licensed data. The company positions the release as a foundation for commercial audio applications, though specific technical specifications have not yet been disclosed.

Google DeepMind released the Gemma 4 E4B assistant model using Multi-Token Prediction (MTP) architecture that accelerates inference by up to 2x through speculative decoding. The 4.5B effective parameter model supports 128K context windows and handles text, image, and audio input with pricing not yet disclosed.

Google DeepMind has released a Multi-Token Prediction assistant model for Gemma 4 26B A4B that achieves up to 2x decoding speedup through speculative decoding. The model uses 3.8B active parameters from a 25.2B total parameter MoE architecture with 128 experts and a 256K token context window.

Google DeepMind has released Gemma 4, a family of open-weight multimodal models including a 31B dense model with 256K context window and four size variants ranging from 2.3B to 30.7B effective parameters. The release includes Multi-Token Prediction (MTP) draft models that achieve up to 2x decoding speedup through speculative decoding while maintaining identical output quality.

Mistral AI released Mistral Medium 3.5, a 128B parameter dense model with a 256k context window that unifies instruction-following, reasoning, and coding capabilities. The model features configurable reasoning effort per request and a vision encoder trained from scratch for variable image sizes.

DeepSeek released V4 in two versions: a 284 billion parameter Flash model and a 1.6 trillion parameter Pro model with 49 billion active parameters. According to DeepSeek, the models use 9.5x-13.7x less memory than V3 through compressed attention mechanisms and FP4/FP8 mixed precision, while supporting a 1 million token context window.

DeepSeek released two preview models: V4 Pro (1.6T total parameters, 49B active) and V4 Flash (284B total, 13B active), both with 1 million token context windows. V4 Pro is priced at $1.74/M input tokens and $3.48/M output—42% cheaper than Claude Sonnet 4.6—while V4 Flash at $0.14/$0.28 per million tokens undercuts all small frontier models.

Alibaba Qwen released Qwen3.6-27B, a 27B parameter dense model that claims flagship-level coding performance surpassing their previous 397B MoE model across major coding benchmarks. The full model is 55.6GB compared to 807GB for the predecessor.

Alibaba Qwen released Qwen3.6-27B, a 27-billion parameter model with a 262,144 token context window extensible to 1,010,000 tokens. The model claims 77.2% on SWE-bench Verified and 53.5% on SWE-bench Pro, with open weights available on Hugging Face.

Open-weight models from Google, Alibaba, Microsoft, and Nvidia have crossed a threshold from research projects to enterprise-grade systems. The shift reflects a growing divide: frontier models from OpenAI and Anthropic are too expensive and pose data security risks for most enterprises, while open alternatives now deliver sufficient capability at a fraction of the cost.

Google DeepMind released Gemma 4, an open-weights multimodal model family in four sizes (2.3B to 31B parameters) with context windows up to 256K tokens. All models support text and image input, with audio native to E2B and E4B variants. The Gemma 4 31B dense model scores 85.2% on MMLU Pro, 89.2% on AIME 2026, and 80.0% on LiveCodeBench—significant improvements over Gemma 3.

Z.ai has released GLM-5.1, a 754-billion parameter open-weight model matching the size of its predecessor GLM-5. The model demonstrates improved ability to generate complex, multi-part outputs like HTML pages with SVG graphics and CSS animations, available via Hugging Face and OpenRouter.

Google DeepMind released the Gemma 4 family of open-weights multimodal models in four sizes: E2B (2.3B effective parameters), E4B (4.5B effective), 26B A4B (3.8B active parameters), and 31B dense. All models support text and image input with 128K-256K context windows; E2B and E4B add native audio capabilities. Models feature reasoning modes, function calling, and multilingual support across 140+ languages.

Google DeepMind released Gemma 4, an open-weights multimodal model family in four sizes (E2B, E4B, 26B A4B, 31B) with context windows up to 256K tokens and native reasoning capabilities. The 26B A4B variant uses Mixture-of-Experts architecture with 3.8B active parameters for efficient inference. All models support text, image input and handle 140+ languages with Apache 2.0 licensing.

Google released Gemma 4, a new line of open-weights models available in sizes from 2 billion to 31 billion parameters, under a permissive Apache 2.0 license. The release includes multimodal capabilities, support for 140+ languages, native function calling, and a 256,000-token context window for the larger variants.

Google DeepMind released Gemma 4, a family of open-weights multimodal models in four sizes: E2B (2.3B effective), E4B (4.5B effective), 26B A4B (3.8B active), and 31B (30.7B parameters). All models support text and image input with 128K-256K context windows, while E2B and E4B add native audio capabilities and reasoning modes across 140+ languages.

Google DeepMind released the Gemma 4 family of open-weights multimodal models in four sizes: E2B (2.3B effective), E4B (4.5B effective), 26B A4B (25.2B total, 3.8B active), and 31B dense. All models support text and image input with 128K-256K context windows, reasoning modes, and native function calling for agentic workflows.

Google DeepMind released the Gemma 4 family of open-weights models ranging from 2.3B to 31B parameters, featuring up to 256K token context windows and native support for text, image, video, and audio inputs. The flagship 31B model scores 85.2% on MMLU Pro and 89.2% on AIME 2026, with a smaller 26B MoE variant requiring only 3.8B active parameters for faster inference.

Mistral has released Voxtral TTS, a 4-billion-parameter text-to-speech model that can clone voices from just three seconds of reference audio across nine languages. The model delivers 70ms latency for typical 10-second samples and outperformed ElevenLabs Flash v2.5 in naturalness tests. Voxtral is available via API at $0.016 per 1,000 characters and as open-weights on Hugging Face.

Mistral AI released Voxtral-4B-TTS-2603, an open-weights text-to-speech model designed for production voice agents. The 4B-parameter model supports 9 languages, 20 preset voices, achieves 70ms latency at concurrency 1 on a single NVIDIA H200, and requires only 16GB GPU memory.

NVIDIA Nemotron 3 Super, a hybrid Mixture of Experts model with 120B parameters and 12B active parameters, is now available as a fully managed model on Amazon Bedrock. The model supports up to 256K token context length and claims 5x higher throughput efficiency over the previous Nemotron Super and 2x higher accuracy on reasoning tasks.

Mistral has released Leanstral, a 120B-parameter code verification agent built with the Lean programming language, claiming it outperforms larger open-source models and offers significant cost advantages over Anthropic's Claude suite. The model achieves a pass@2 score of 26.3—beating Claude Sonnet by 2.6 points—while costing $36 to run compared to Sonnet's $549.