diff --git a/assets/amazon.yaml b/assets/amazon.yaml index d539cd24..70aad8fb 100644 --- a/assets/amazon.yaml +++ b/assets/amazon.yaml @@ -86,3 +86,76 @@ prohibited_uses: '' monitoring: '' feedback: https://github.com/amazon-science/chronos-forecasting/discussions +- type: model + name: Amazon Nova (Understanding) + organization: Amazon Web Services (AWS) + description: A new generation of state-of-the-art foundation models (FMs) that + deliver frontier intelligence and industry leading price performance, available + exclusively in Amazon Bedrock. Amazon Nova understanding models excel in Retrieval-Augmented + Generation (RAG), function calling, and agentic applications. + created_date: 2024-12-03 + url: https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/ + model_card: unknown + modality: + explanation: Amazon Nova understanding models accept text, image, or video inputs + to generate text output. + value: text, image, video; text + analysis: Amazon Nova Pro is capable of processing up to 300K input tokens and + sets new standards in multimodal intelligence and agentic workflows that require + calling APIs and tools to complete complex workflows. It achieves state-of-the-art + performance on key benchmarks including visual question answering ( TextVQA + ) and video understanding ( VATEX ). + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: All Amazon Nova models include built-in safety controls and creative + content generation models include watermarking capabilities to promote responsible + AI use. + access: + explanation: available exclusively in Amazon Bedrock + value: limited + license: unknown + intended_uses: You can build on Amazon Nova to analyze complex documents and videos, + understand charts and diagrams, generate engaging video content, and build sophisticated + AI agents, from across a range of intelligence classes optimized for enterprise + workloads. + prohibited_uses: unknown + monitoring: unknown + feedback: unknown +- type: model + name: Amazon Nova (Creative Content Generation) + organization: Amazon Web Services (AWS) + description: A new generation of state-of-the-art foundation models (FMs) that + deliver frontier intelligence and industry leading price performance, available + exclusively in Amazon Bedrock. + created_date: 2024-12-03 + url: https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/ + model_card: unknown + modality: + explanation: Amazon creative content generation models accept text and image + inputs to generate image or video output. + value: text, image;image, video + analysis: Amazon Nova Canvas excels on human evaluations and key benchmarks such + as text-to-image faithfulness evaluation with question answering (TIFA) and + ImageReward. + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: All Amazon Nova models include built-in safety controls and creative + content generation models include watermarking capabilities to promote responsible + AI use. + access: + explanation: available exclusively in Amazon Bedrock + value: limited + license: unknown + intended_uses: You can build on Amazon Nova to analyze complex documents and videos, + understand charts and diagrams, generate engaging video content, and build sophisticated + AI agents, from across a range of intelligence classes optimized for enterprise + workloads. + prohibited_uses: unknown + monitoring: unknown + feedback: unknown diff --git a/assets/anthropic.yaml b/assets/anthropic.yaml index 1fe22119..758ed030 100644 --- a/assets/anthropic.yaml +++ b/assets/anthropic.yaml @@ -608,15 +608,17 @@ speed of its predecessor, Claude 3 Opus, and is designed to tackle tasks like context-sensitive customer support, orchestrating multi-step workflows, interpreting charts and graphs, and transcribing text from images. - created_date: 2024-06-21 - url: https://www.anthropic.com/news/claude-3-5-sonnet + created_date: + explanation: Claude 3.5 Sonnet updated on Oct. 22, initially released on June + 20 of the same year. + value: 2024-10-22 + url: https://www.anthropic.com/news/3-5-models-and-computer-use model_card: unknown modality: text; image, text analysis: The model has been evaluated on a range of tests including graduate-level reasoning (GPQA), undergraduate-level knowledge (MMLU), coding proficiency (HumanEval), - and standard vision benchmarks. In an internal agentic coding evaluation, Claude - 3.5 Sonnet solved 64% of problems, outperforming the previous version, Claude - 3 Opus, which solved 38%. + and standard vision benchmarks. Claude 3.5 Sonnet demonstrates state-of-the-art + performance on most benchmarks. size: Unknown dependencies: [] training_emissions: Unknown @@ -637,3 +639,38 @@ integrated to ensure robustness of evaluations. feedback: Feedback on Claude 3.5 Sonnet can be submitted directly in-product to inform the development roadmap and improve user experience. +- type: model + name: Claude 3.5 Haiku + organization: Anthropic + description: Claude 3.5 Haiku is Anthropic's fastest model, delivering advanced + coding, tool use, and reasoning capability, surpassing the previous Claude 3 + Opus in intelligence benchmarks. It is designed for critical use cases where + low latency is essential, such as user-facing chatbots and code completions. + created_date: 2024-10-22 + url: https://www.anthropic.com/claude/haiku + model_card: unknown + modality: + explanation: Claude 3.5 Haiku is available...initially as a text-only model + and with image input to follow. + value: text; unknown + analysis: Claude 3.5 Haiku offers strong performance and speed across a variety + of coding, tool use, and reasoning tasks. Also, it has been tested in extensive + safety evaluations and exceeded expectations in reasoning and code generation + tasks. + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: During Claude 3.5 Haiku’s development, we conducted extensive + safety evaluations spanning multiple languages and policy domains. + access: + explanation: Claude 3.5 Haiku is available across Claude.ai, our first-party + API, Amazon Bedrock, and Google Cloud’s Vertex AI. + value: open + license: unknown + intended_uses: Critical use cases where low latency matters, like user-facing + chatbots and code completions. + prohibited_uses: unknown + monitoring: unknown + feedback: unknown diff --git a/assets/cohere.yaml b/assets/cohere.yaml index bc7e0a2f..c5f86d73 100644 --- a/assets/cohere.yaml +++ b/assets/cohere.yaml @@ -592,3 +592,25 @@ prohibited_uses: unknown monitoring: unknown feedback: https://huggingface.co/CohereForAI/aya-23-35B/discussions +- type: model + name: Command R+ + organization: Cohere + description: Command R+ is a state-of-the-art RAG-optimized model designed to + tackle enterprise-grade workloads, and is available first on Microsoft Azure. + created_date: 2024-04-04 + url: https://cohere.com/blog/command-r-plus-microsoft-azure + model_card: unknown + modality: unknown + analysis: unknown + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: unknown + access: '' + license: unknown + intended_uses: unknown + prohibited_uses: unknown + monitoring: unknown + feedback: unknown diff --git a/assets/genmo.yaml b/assets/genmo.yaml new file mode 100644 index 00000000..7425032a --- /dev/null +++ b/assets/genmo.yaml @@ -0,0 +1,43 @@ +--- +- type: model + name: Mochi 1 + organization: Genmo + description: Mochi 1 is an open-source video generation model designed to produce + high-fidelity motion and strong prompt adherence in generated videos, setting + a new standard for open video generation systems. + created_date: 2025-01-14 + url: https://www.genmo.ai/blog + model_card: unknown + modality: + explanation: Mochi 1 generates smooth videos... Measures how accurately generated + videos follow the provided textual instructions + value: text; video + analysis: Mochi 1 sets a new best-in-class standard for open-source video generation. + It also performs very competitively with the leading closed models... We benchmark + prompt adherence with an automated metric using a vision language model as a + judge following the protocol in OpenAI DALL-E 3. We evaluate generated videos + using Gemini-1.5-Pro-002. + size: + explanation: featuring a 10 billion parameter diffusion model + value: 10B parameters + dependencies: [DDPM, DreamFusion, Emu Video, T5-XXL] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: robust safety moderation protocols in the playground to ensure + that all video generations remain safe and aligned with ethical guidelines. + access: + explanation: open state-of-the-art video generation model... The weights and + architecture for Mochi 1 are open + value: open + license: + explanation: We're releasing the model under a permissive Apache 2.0 license. + value: Apache 2.0 + intended_uses: Advance the field of video generation and explore new methodologies. + Build innovative applications in entertainment, advertising, education, and + more. Empower artists and creators to bring their visions to life with AI-generated + videos. Generate synthetic data for training AI models in robotics, autonomous + vehicles and virtual environments. + prohibited_uses: unknown + monitoring: unknown + feedback: unknown diff --git a/assets/google.yaml b/assets/google.yaml index 8bc851ae..94d73f70 100644 --- a/assets/google.yaml +++ b/assets/google.yaml @@ -1908,3 +1908,69 @@ monitoring: unknown feedback: Encourages developer feedback to inform model improvements and future updates. +- type: model + name: Veo 2 + organization: Google DeepMind + description: Veo 2 is a state-of-the-art video generation model that creates videos + with realistic motion and high-quality output, up to 4K, with extensive camera + controls. It simulates real-world physics and offers advanced motion capabilities + with enhanced realism and fidelity. + created_date: 2024-12-16 + url: https://deepmind.google/technologies/veo/veo-2/ + model_card: unknown + modality: + explanation: Our state-of-the-art video generation model ... text-to-image model + Veo 2 + value: text; video + analysis: Veo 2 outperforms other leading video generation models, based on human + evaluations of its performance. + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: Veo 2 includes features that enhance realism, fidelity, detail, + and artifact reduction to ensure high-quality output. + access: limited + license: unknown + intended_uses: Creating high-quality videos with realistic motion, different styles, + camera controls, shot styles, angles, and movements. + prohibited_uses: unknown + monitoring: unknown + feedback: unknown + +- type: model + name: Gemini 2.0 + organization: Google DeepMind + description: Google DeepMind introduces Gemini 2.0, a new AI model designed for + the 'agentic era.' + created_date: 2024-12-11 + url: https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/#ceo-message + model_card: unknown + modality: + explanation: The first model built to be natively multimodal, Gemini 1.0 and + 1.5 drove big advances with multimodality and long context to understand information + across text, video, images, audio and code... + value: text, video, image, audio; image, text + analysis: unknown + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: + explanation: It’s built on custom hardware like Trillium, our sixth-generation + TPUs. + value: custom hardware like Trillium, our sixth-generation TPUs + quality_control: Google is committed to building AI responsibly, with safety and + security as key priorities. + access: + explanation: Gemini 2.0 Flash is available to developers and trusted testers, + with wider availability planned for early next year. + value: limited + license: unknown + intended_uses: Develop more agentic models, meaning they can understand more about + the world around you, think multiple steps ahead, and take action on your behalf, + with your supervision. + prohibited_uses: unknown + monitoring: unknown + feedback: unknown diff --git a/assets/ibm.yaml b/assets/ibm.yaml index 260b3a9f..368680ae 100644 --- a/assets/ibm.yaml +++ b/assets/ibm.yaml @@ -75,3 +75,45 @@ prohibited_uses: '' monitoring: '' feedback: '' +- type: model + name: IBM Granite 3.0 + organization: IBM + description: IBM Granite 3.0 models deliver state-of-the-art performance relative + to model size while maximizing safety, speed and cost-efficiency for enterprise + use cases. + created_date: 2024-10-21 + url: https://www.ibm.com/new/ibm-granite-3-0-open-state-of-the-art-enterprise-models + model_card: unknown + modality: + explanation: IBM Granite 3.0 8B Instruct model for classic natural language + use cases including text generation, classification, summarization, entity + extraction and customer service chatbots + value: text; text + analysis: Granite 3.0 8B Instruct matches leading similarly-sized open models + on academic benchmarks while outperforming those peers on benchmarks for enterprise + tasks and safety. + size: + explanation: 'Dense, general purpose LLMs: Granite-3.0-8B-Instruct' + value: 8B parameters + dependencies: [Hugging Face’s OpenLLM Leaderboard v2] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: The entire Granite family of models are trained on carefully + curated enterprise datasets, filtered for objectionable content with critical + concerns like governance, risk, privacy and bias mitigation in mind + access: + explanation: In keeping with IBM’s strong historical commitment to open source + , all Granite models are released under the permissive Apache 2.0 license + value: open + license: + explanation: In keeping with IBM’s strong historical commitment to open source + , all Granite models are released under the permissive Apache 2.0 license + value: Apache 2.0 + intended_uses: classic natural language use cases including text generation, classification, + summarization, entity extraction and customer service chatbots, programming + language use cases such as code generation, code explanation and code editing, + and for agentic use cases requiring tool calling + prohibited_uses: unknown + monitoring: '' + feedback: unknown diff --git a/assets/inflection.yaml b/assets/inflection.yaml index 4f82b1f9..d18db5ce 100644 --- a/assets/inflection.yaml +++ b/assets/inflection.yaml @@ -93,3 +93,29 @@ prohibited_uses: '' monitoring: '' feedback: none +- type: model + name: Inflection 3.0 + organization: Inflection AI + description: Inflection for Enterprise, powered by our industry-first, enterprise-grade + AI system, Inflection 3.0. + created_date: 2024-10-07 + url: https://inflection.ai/blog/enterprise + model_card: unknown + modality: unknown + analysis: unknown + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: unknown + access: + explanation: Developers can now access Inflection AI’s Large Language Model + through its new commercial API. + value: open + license: unknown + intended_uses: unknown + prohibited_uses: unknown + monitoring: unknown + feedback: So please drop us a line. We want to keep hearing from enterprises about + how we can help solve their challenges and make AI a reality for their business. diff --git a/assets/meta.yaml b/assets/meta.yaml index 0b7653ec..8b337691 100644 --- a/assets/meta.yaml +++ b/assets/meta.yaml @@ -891,3 +891,51 @@ prohibited_uses: Unknown monitoring: Unknown feedback: Unknown +- type: model + name: Llama 3.3 + organization: Meta + description: The Meta Llama 3.3 multilingual large language model (LLM) is an + instruction tuned generative model in 70B (text in/text out). + created_date: 2024-12-06 + url: https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct + model_card: https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct + modality: + explanation: The Llama 3.3 instruction tuned text only model is optimized for + multilingual dialogue use cases. + value: text; text + analysis: Unknown + size: + explanation: The Meta Llama 3.3 multilingual large language model (LLM) is an + instruction tuned generative model in 70B (text in/text out). + value: 70B parameters + dependencies: [] + training_emissions: + explanation: Training Greenhouse Gas Emissions Estimated total location-based + greenhouse gas emissions were 11,390 tons CO2eq for training. + value: 11,390 tons CO2eq + training_time: + explanation: Training utilized a cumulative of 39.3M GPU hours of computation + on H100-80GB (TDP of 700W) type hardware. + value: 39.3M GPU hours + training_hardware: + explanation: Training utilized a cumulative of 39.3M GPU hours of computation + on H100-80GB (TDP of 700W) type hardware. + value: H100-80GB (TDP of 700W) type hardware + quality_control: Used "supervised fine-tuning (SFT) and reinforcement learning + with human feedback (RLHF) to align with human preferences for helpfulness and + safety." + access: + explanation: Future versions of the tuned models will be released as we improve + model safety with community feedback. + value: open + license: + explanation: A custom commercial license, the Llama 3.3 Community License Agreement + value: Llama 3.3 Community License Agreement + intended_uses: Intended for commercial and research use in multiple languages. + Instruction tuned text only models are intended for assistant-like chat. + prohibited_uses: Use in any manner that violates applicable laws or regulations + (including trade compliance laws). Use in any other way that is prohibited by + the Acceptable Use Policy and Llama 3.3 Community License. + monitoring: Unknown + feedback: Instructions on how to provide feedback or comments on the model can + be found in the model README. diff --git a/assets/microsoft.yaml b/assets/microsoft.yaml index 137996a4..1c7dc960 100644 --- a/assets/microsoft.yaml +++ b/assets/microsoft.yaml @@ -1032,3 +1032,38 @@ use case, particularly for high risk scenarios. monitoring: Unknown feedback: Unknown +- type: model + name: Phi-4 + organization: Microsoft + description: the latest small language model in Phi family, that offers high quality + results at a small size (14B parameters). + created_date: 2024-12-13 + url: https://techcommunity.microsoft.com/blog/aiplatformblog/introducing-phi-4-microsoft%E2%80%99s-newest-small-language-model-specializing-in-comple/4357090 + model_card: unknown + modality: + explanation: Today we are introducing Phi-4 , our 14B parameter state-of-the-art + small language model (SLM) that excels at complex reasoning in areas such + as math, in addition to conventional language processing. + value: text; text + analysis: Phi-4 outperforms comparable and larger models on math related reasoning. + size: + explanation: a small size (14B parameters). + value: 14B parameters + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: Building AI solutions responsibly is at the core of AI development + at Microsoft. We have made our robust responsible AI capabilities available + to customers building with Phi models. + access: + explanation: Phi-4 is available on Azure AI Foundry and on Hugging Face. + value: open + license: unknown + intended_uses: Specialized in complex reasoning, particularly good at math problems + and high-quality language processing. + prohibited_uses: unknown + monitoring: Azure AI evaluations in AI Foundry enable developers to iteratively + assess the quality and safety of models and applications using built-in and + custom metrics to inform mitigations. + feedback: unknown diff --git a/assets/mistral.yaml b/assets/mistral.yaml index 3c78ac23..7fe7ce5a 100644 --- a/assets/mistral.yaml +++ b/assets/mistral.yaml @@ -192,3 +192,77 @@ monitoring: Unknown feedback: Feedback is likely expected to be given through the HuggingFace platform where the model's weights are hosted or directly to the Mistral AI team. +- type: model + name: Pixtral Large + organization: Mistral AI + description: Pixtral Large is the second model in our multimodal family and demonstrates + frontier-level image understanding. Particularly, the model is able to understand + documents, charts and natural images, while maintaining the leading text-only + understanding of Mistral Large 2. + created_date: 2024-11-18 + url: https://mistral.ai/news/pixtral-large/ + model_card: unknown + modality: + explanation: Pixtral Large is the second model in our multimodal family and + demonstrates frontier-level image understanding. + value: text, image; text + analysis: We evaluate Pixtral Large against frontier models on a set of standard + multimodal benchmarks, through a common testing harness. + size: + explanation: Today we announce Pixtral Large, a 124B open-weights multimodal + model. + value: 124B parameters + dependencies: [Mistral Large 2] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: unknown + access: + explanation: The model is available under the Mistral Research License (MRL) + for research and educational use; and the Mistral Commercial License for experimentation, + testing, and production for commercial purposes. + value: open + license: + explanation: The model is available under the Mistral Research License (MRL) + for research and educational use; and the Mistral Commercial License for experimentation, + testing, and production for commercial purposes. + value: Mistral Research License (MRL), Mistral Commercial License + intended_uses: RAG and agentic workflows, making it a suitable choice for enterprise + use cases such as knowledge exploration and sharing, semantic understanding + of documents, task automation, and improved customer experiences. + prohibited_uses: unknown + monitoring: unknown + feedback: unknown + +- type: model + name: Codestral 25.01 + organization: Mistral AI + description: Lightweight, fast, and proficient in over 80 programming languages, + Codestral is optimized for low-latency, high-frequency usecases and supports + tasks such as fill-in-the-middle (FIM), code correction and test generation. + created_date: 2025-01-13 + url: https://mistral.ai/news/codestral-2501/ + model_card: unknown + modality: + explanation: it for free in Continue for VS Code or JetBrains + value: text; text + analysis: Benchmarks We have benchmarked the new Codestral with the leading sub-100B + parameter coding models that are widely considered to be best-in-class for FIM + tasks. + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: unknown + access: + explanation: The API is also available on Google Cloud’s Vertex AI, in private + preview on Azure AI Foundry, and coming soon to Amazon Bedrock. + value: closed + license: unknown + intended_uses: Highly capable coding companion, regularly boosting productivity + several times over. + prohibited_uses: unknown + monitoring: unknown + feedback: We can’t wait to hear your experience! Try it now Try it on Continue.dev + with VsCode or JetBrains diff --git a/assets/openx.yaml b/assets/openx.yaml index 25d61646..afaf3ab6 100644 --- a/assets/openx.yaml +++ b/assets/openx.yaml @@ -77,3 +77,79 @@ prohibited_uses: none monitoring: unknown feedback: none +- type: model + name: GPT-4o + organization: OpenAI + description: GPT-4o is an autoregressive omni model that accepts a combination + of text, audio, image, and video as input and produces any combination of text, + audio, and image outputs. It is trained end-to-end across text, vision, and + audio, focusing on multimodal capabilities. + created_date: 2024-08-08 + url: https://arxiv.org/pdf/2410.21276 + model_card: unknown + modality: + explanation: '...accepts as input any combination of text, audio, image, and + video and generates any combination of text, audio, and image outputs.' + value: text, audio, image, video; text, audio, image + analysis: GPT-4o underwent evaluations that included the Preparedness Framework, + external red teaming, and third-party assessments to ensure safe and aligned + deployment. The evaluations focused on identifying and mitigating potential + risks across its capabilities, especially speech-to-speech functionality. + size: unknown + dependencies: [Shutterstock] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: Quality and safety measures included prior risk assessments, + post-training mitigation, moderation tools, advanced data filtering, and external + red teaming efforts with experts to evaluate potential risks like bias, discrimination, + and information harms. + access: + explanation: we are sharing the GPT-4o System Card, which includes our Preparedness + Framework evaluations. + value: limited + license: unknown + intended_uses: Use in multimodal applications requiring understanding and generation + of combinations of text, audio, and image outputs, better performance on non-English + languages, and enhanced vision and audio understanding. + prohibited_uses: Uses that could involve bias, discrimination, harmful content, + or violation of usage policies. + monitoring: Continuous monitoring and enforcement, providing moderation tools + and transparency reports, and gathering feedback from users. + feedback: unknown + +- type: model + name: OpenAI o1 + organization: OpenAI + description: OpenAI o1 is a new series of AI models designed to spend more time + thinking before they respond. They can reason through complex tasks and solve + harder problems than previous models in science, coding, and math. + created_date: 2024-09-12 + url: https://openai.com/o1/ + model_card: unknown + modality: text; text + analysis: Evaluated on challenging benchmark tasks in physics, chemistry, and + biology. In a qualifying exam for the International Mathematics Olympiad (IMO), + GPT-4o correctly solved only 13% of problems, while the reasoning model o1 scored + 83%. + size: unknown + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: To match the new capabilities of these models, OpenAI has bolstered + safety work, internal governance, and federal government collaboration. This + includes rigorous testing and evaluations using their Preparedness Framework⁠(opens + in a new window), best-in-class red teaming, and board-level review processes, + including by OpenAI's Safety & Security Committee. + access: limited + license: unknown + intended_uses: These enhanced reasoning capabilities may be particularly useful + if you’re tackling complex problems in science, coding, math, and similar fields. + For example, o1 can be used by healthcare researchers to annotate cell sequencing + data, by physicists to generate complicated mathematical formulas needed for + quantum optics, and by developers in all fields to build and execute multi-step + workflows. + prohibited_uses: '' + monitoring: '' + feedback: unknown diff --git a/assets/stability_ai.yaml b/assets/stability_ai.yaml index 3ff8188c..0972b1bb 100644 --- a/assets/stability_ai.yaml +++ b/assets/stability_ai.yaml @@ -113,3 +113,43 @@ monitoring: Unknown feedback: Information on any downstream issues with the model can be reported to Stability AI through their support request system. +- type: model + name: Stable Diffusion 3.5 + organization: Stability AI + description: Stable Diffusion 3.5 reflects our commitment to empower builders + and creators with tools that are widely accessible, cutting-edge, and free for + most use cases. + created_date: 2023-10-29 + url: https://stability.ai/news/introducing-stable-diffusion-3-5 + model_card: unknown + modality: + explanation: Capable of generating a wide range of styles and aesthetics like + 3D, photography, painting, line art, and virtually any visual style imaginable. + value: text; image + analysis: Our analysis shows that Stable Diffusion 3.5 Large leads the market + in prompt adherence and rivals much larger models in image quality. + size: + explanation: At 8.1 billion parameters, with superior quality and prompt adherence, + this base model is the most powerful in the Stable Diffusion family. + value: 8.1B parameters + dependencies: [] + training_emissions: unknown + training_time: unknown + training_hardware: unknown + quality_control: We believe in safe, responsible AI practices and take deliberate + measures to ensure Integrity starts at the early stages of development. + access: + explanation: This open release includes multiple model variants, including Stable + Diffusion 3.5 Large and Stable Diffusion 3.5 Large Turbo, and as of October + 29th, Stable Diffusion 3.5 Medium. + value: open + license: + explanation: This open release includes multiple variants that are customizable, + run on consumer hardware, and are available for use under the permissive Stability + AI Community License. + value: Stability AI Community + intended_uses: This model is ideal for professional use cases at 1 megapixel resolution. + prohibited_uses: unknown + monitoring: unknown + feedback: We look forward to hearing your feedback on Stable Diffusion 3.5 and + seeing what you create with the models. diff --git a/assets/unknown.yaml b/assets/unknown.yaml new file mode 100644 index 00000000..c2f323c3 --- /dev/null +++ b/assets/unknown.yaml @@ -0,0 +1,40 @@ +--- +- type: model + name: DeepSeek-V3 + organization: DeepSeek + description: DeepSeek-V3 is a Mixture-of-Experts (MoE) language model with 671B + total parameters and 37B activated per token. It utilizes Multi-head Latent + Attention (MLA) and adopts innovative strategies for improved performance, such + as an auxiliary-loss-free load balancing and a multi-token prediction training + objective. Comprehensive evaluations show it achieves performance comparable + to leading closed-source models. + created_date: 2025-01-14 + url: https://huggingface.co/deepseek-ai/DeepSeek-V3 + model_card: https://huggingface.co/deepseek-ai/DeepSeek-V3 + modality: unknown + analysis: Comprehensive evaluations reveal that DeepSeek-V3 outperforms other + open-source models and achieves performance comparable to leading closed-source + models. + size: + explanation: a strong Mixture-of-Experts (MoE) language model with 671B total + parameters with 37B activated for each token. + value: 671B parameters (sparse) + dependencies: [DeepSeek-R1] + training_emissions: unknown + training_time: + explanation: DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. + value: 2.788M GPU hours + training_hardware: + explanation: DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. + value: H800 GPUs + quality_control: Post-training includes knowledge distillation from the DeepSeek-R1 + model, incorporating verification and reflection patterns to enhance reasoning + performance. + access: + explanation: producing the currently strongest open-source base model. + value: open + license: MIT + intended_uses: unknown + prohibited_uses: unknown + monitoring: unknown + feedback: unknown diff --git a/js/main.js b/js/main.js index 6b7611e5..3809eae1 100644 --- a/js/main.js +++ b/js/main.js @@ -635,9 +635,11 @@ function loadAssetsAndRenderPageContent() { const paths = [ 'assets/adept.yaml', + 'assets/genmo.yaml', 'assets/mila.yaml', 'assets/soochow.yaml', 'assets/baichuan.yaml', + 'assets/unknown.yaml', 'assets/xwin.yaml', 'assets/mistral.yaml', 'assets/adobe.yaml',