Alternatives to d1.2

Compare d1.2 alternatives for your business or organization using the curated list below. SourceForge ranks the best alternatives to d1.2 in 2026. Compare features, ratings, user reviews, pricing, and more from d1.2 competitors and alternatives in order to make an informed decision for your business.

  • 1
    GPT-6 Luna
    GPT-6 Luna is OpenAI’s cost-efficient GPT-6 model designed for everyday professional work, coding, computer use, and agentic applications at scale. It brings many of the advances introduced with GPT-6 Astra to a faster and substantially lower-cost model while improving on GPT-5.6 Luna in capability and factual reliability. The model supports adjustable reasoning effort, allowing applications to spend more compute on harder tasks and less on simpler requests. GPT-6 Luna can handle software engineering, multi-step business workflows, computer interaction, and other tool-using tasks that benefit from low operating cost. Improved GPT-6 prompt caching helps long-running agents reuse more context, respond faster, and reduce the cost of repeated input. GPT-6 Luna is available in ChatGPT Work, Codex, the ChatGPT desktop app for Free and Go users, and the OpenAI API as gpt-6-luna.
    5.0 out of 5 stars 1 Rating for GPT-6 Luna
    Starting Price: $0.10 per 1M tokens (input)
  • 2
    MiniMax H3

    MiniMax H3

    MiniMax

    MiniMax H3 is a general-purpose omni-modal generation model that jointly understands multimodal contexts spanning text, images, video, and audio. It generates videos with native stereo sound at up to 2K resolution and 15 seconds in length, delivering content for advertising, branding, ecommerce, product design, UI/UX, gaming, and creative workflows. Users can combine reference types in one instruction, for example, transferring camera movement from a video, placing a character from an image into the scene, and matching vocals from an audio clip, while describing the relationships in natural language. H3 supports text-to-image, text-to-video with jointly generated audio, multi-shot modeling, text-to-audio, and generalized reference and editing across images, videos, and audio. Voice, sound effects, and music are modeled together. The model excels at instruction following, accurate text and brand presentation, and video-to-video motion transfer.
  • 3
    Jev

    Jev

    TypeSafe AI

    Jev is a System One Model from TypeSafe AI designed for fast, structured decision-making inside software and automated workflows. Unlike conventional large language models that generate free-form text token by token, Jev accepts unstructured state and returns predefined type-safe values with calibrated probabilities and confidence scores. The model uses a parallel sampling architecture and a training approach called Reinforcement Learning for Calibrated Decisions, which is optimized for producing reliable probabilistic decisions rather than conversational responses. Jev is intended for tasks such as classification, routing, scoring, extraction, branching logic, verification, guardrails, and other AI-powered workflows where structured outputs can plug directly into application code. TypeSafe reports end-to-end response times ranging from about 70 to 500 milliseconds and says Jev can be substantially faster and more efficient than frontier language models on System One-style workloads.
    Starting Price: Input: $0.042 / 1M tokens
  • 4
    Perplexity Decider
    Perplexity Decider is a 27B decision model fine-tuned from Qwen3.8-27B for structured classification and multimodal decision tasks. Rather than generating unrestricted text, it takes an input state and a typed decision schema, then returns the selected result together with calibrated probabilities. The model supports choice decisions for selecting among predefined categories and noul decisions for estimating the probability of a yes-or-no condition. It can also accept images alongside text, allowing applications to make decisions from screenshots or other visual inputs. Example workflows include routing a support request to billing, technical support, or sales, and estimating whether a message expresses urgency. Perplexity reports an overall accuracy of 85.71% across 11 benchmarks, including FinancialPhraseBank, RAGTruth, JudgeBench, TabFact, ContractNLI, Circa, Belebele, and TruthfulQA binary.
  • 5
    OpenAI Decisions API
    The Decisions API turns text and images into typed decisions an application can use, evaluating inputs about 10x faster than the Responses API. Developers provide a model, shared input evidence, and one or more questions describing what should be evaluated, and the API returns structured answers that can be used to classify content, route requests, prioritize work, or make threshold-based decisions. It supports three question types: predicate checks whether a condition is true and returns a probability from 0 to 1; choice selects one value from a fixed set and returns option probabilities and confidence; and score evaluates an input against ordered rubric levels and returns a probability-weighted score. Inputs can be plain text or user messages containing both text and images, allowing applications to evaluate cases such as visible product damage, request intent, document type, issue severity, or answer relevance.
  • 6
    d1.2-ultra

    d1.2-ultra

    Liquid AI

    d1.2-ultra is an advanced multimodal artificial intelligence decision model developed by Liquid AI for classification, information extraction, visual analysis, and probabilistic decision-making. The model processes text, images, and audio to answer structured questions without generating lengthy conversational responses. Its capabilities include extracting text, numbers, and dates from documents, identifying sensitive information, locating objects in images, and estimating numerical quantities. d1.2-ultra also supports Boolean decisions, multiple-choice classifications, multi-label selections, and numerical scoring with calibrated confidence probabilities. The model achieved a score of 75 on the Artificial Analysis Decision Index v2.1, ranking first among evaluated decision models at the time of its release.
    Starting Price: $0.042 per 1M tokens (input)
  • 7
    SubQ Snap 1.1

    SubQ Snap 1.1

    Subquadratic

    SubQ Snap is a hosted multimodal decision model for developers building AI agents and enterprise workflows. Through Subquadratic’s API, applications send text or image context with defined questions and receive structured answers such as labels, choices, scores, and probabilities. This supports tasks including tool and file selection, routing, triage, guardrails, ranking, classification, and checking whether a test or result meets specified criteria. The model accepts up to 384K tokens of context and supports image input. It is designed for bounded decisions rather than open-ended text generation, so it can be used alongside a language model that handles planning or writing. Public-preview API documentation includes request examples, model access details, and decision formats.
  • 8
    Gemini Decision 1
    Gemini Decision 1 is an artificial intelligence decision-making model developed by Google for classification, routing, scoring, and structured evaluation tasks. Unlike traditional generative AI models that produce conversational responses, Gemini Decision 1 analyzes input information and returns structured decisions with associated probabilities and confidence values. The model supports Boolean questions, multiple-choice selections, and numerical or ordinal scoring to help applications make automated decisions. It can process large amounts of contextual information using a context window of approximately 262,000 tokens. Developers can integrate the model into applications for content categorization, customer support triage, workflow automation, and AI agent decision-making. Available through API integrations, Gemini Decision 1 helps organizations build intelligent systems that evaluate information and produce measurable, structured outputs.
  • 9
    Liquid d1

    Liquid d1

    Liquid AI

    d1 is Liquid AI’s first decision model, built to answer structured questions about a situation with probabilities rather than generating text. It takes unstructured data such as text, images, or both, along with one or more questions, processes them in a single forward pass, and returns probabilities without generating output tokens. d1 supports three question types: noul for yes-or-no decisions, choice for selecting among labels with a probability for each option, and score for placing an input on an ordered scale using probability-weighted levels. Multiple questions can be evaluated against the same state in one request, making it useful for real-time decision workflows. Liquid AI demonstrates d1 across applications such as filtering support tickets, browsing code repositories, organizing documents into smart folders, operating websites, compacting agent context, inspecting industrial parts, and playing games from screenshots.
  • 10
    Hume Audio Decisions API
    Hume Audio Decisions API is a hosted API for developers building voice applications and teams evaluating speech and sound. It accepts an audio recording, optional context, and user-defined questions, then returns probabilities for answers in structured form. Questions can cover observable audio facts, such as the number of speakers or background speech, and listener judgments about tone, naturalness, empathy, clarity, or frustration. Teams can use the results to review support calls, test voice agents and text-to-speech output, label speech corpora, compare recordings, and identify examples for escalation or model training. The API works as a decision layer alongside speech recognition and voice generation, allowing application code to use its scores and probabilities in custom workflows. API access is requested through Hume, and billing is based on audio, question, and context input tokens.
  • 11
    Parallel Decisions API
    Parallel Decisions API is a hosted, web-grounded classification API for developers building software, data-enrichment pipelines, and AI-agent workflows. A request supplies context and one or more bounded questions; the API returns typed answers with probabilities and citations to supporting web sources. Multiple questions can share context in a single request, supporting uses such as company or vendor qualification, lead routing, compliance screening, content filtering, and agent guardrails. By retrieving current web evidence, it can assess facts that may have changed since a model's training data. Developers call the REST endpoint with an API key or access the web_decisions tool through Parallel's MCP server. A hosted playground and documentation support testing and implementation.
  • 12
    Solar Decide
    Solar Decide is Upstage’s structured decision model, serving as a System One endpoint on Solar Mini 4. Instead of generating prose, it takes a state together with one or more typed questions and returns a choice, score, or yes/no answer with calibrated probabilities taken directly from the model. Each question is handled in a single forward pass, making the model suited to fast classification, routing, policy checks, prioritization, and other structured decision tasks. The choice type selects among predefined options and returns the selected option, individual probabilities, and confidence. The score type evaluates an input against ordered criteria and returns a probability-weighted score, legend, probability distribution, and confidence. The noul type returns the probability that a yes-or-no proposition is true. Multiple questions can share the same state in one request, and the state can be provided as text or any JSON value serialized into the prompt.
  • 13
    Microsoft-Decision-1
    Microsoft-Decision-1 is a fast decision-scoring model built for structured outputs that software can immediately act on. It is designed for routing, classification, prioritization, verification, and workflow control across applications, agents, and automated systems. The model is post-trained from Qwen3.5-9B for single-pass decision scoring and, when given a fixed set of answer options, returns a calibrated probability for each one. It supports yes/no, multiple-choice, rating, and rubric-based grading of AI responses and agent actions through a structured API. Microsoft evaluated the model across 36 benchmarks covering routing, ranking, long context, multilingual and out-of-distribution tasks, reasoning, and safety, and reports that it achieved the highest accuracy in that comparison while also delivering very low latency. The model is also designed for robustness, with testing focused on keeping decisions stable when prompts are paraphrased.
  • 14
    Perplexity Decisions API
    Perplexity’s Decisions API is built for fast, structured decisions that software can use directly. Instead of generating free-form text, it reads natural-language text, JSON, or images as shared state and returns typed answers with probabilities. Developers can attach one or more named questions to the same input and choose among three question types. noul handles yes-or-no judgments and returns the probability of yes from 0 to 1. choice selects from developer-defined options and returns the most likely option, the probability of every option, and a confidence score. score evaluates content against an ordered rubric and returns a probability-weighted score, the full probability distribution, and confidence. The API is designed for tasks such as classification, routing, grading against a rubric, severity assessment, sentiment analysis, or any decision that can be compared against a threshold.
  • 15
    Clef

    Clef

    Cloudflare

    Clef is a 27B multimodal decision model that turns a state and a schema of typed questions into structured decisions. It can read state information as text, JSON, images, or video and return a probability for every allowed option of every question. The model supports a 65,536-token context window and vision inputs, allowing developers to evaluate both structured application data and visual content in the same decision workflow. Clef supports three typed question formats: noul for estimating the probability that a proposition is true, choice for selecting among predefined options and returning per-option probabilities, and score for evaluating an input against ordered criteria and producing a probability-weighted score. Applications can submit between 1 and 64 questions in a single request, with answers returned under the same question IDs.
  • 16
    Laya AI

    Laya AI

    Laya AI

    Laya AI is an open source family of non-autoregressive System 1 decision models built for fast, typed decisions instead of free-form text generation. Given a state and structured questions, Laya evaluates them in a single forward pass and returns outputs that application code can use directly. It supports three decision types: choice, which selects among named options and returns probabilities; score, which places an input on an ordered rubric such as urgency or severity; and noul, which estimates the probability that a proposition is true. Laya is designed for workloads such as agent and model routing, support ticket triage, guardrails, intent classification, risk and fraud scoring, workflow escalation, RAG filtering, moderation, and phishing detection. It can route requests by department, score urgency, flag refund or churn signals, detect policy or abuse conditions, and escalate uncertain cases to a larger model or human reviewer.
  • 17
    Strands Decider

    Strands Decider

    Strands Labs

    Strands Decider is a small, fast decision model, or System One model, built for agentic AI workflows. Unlike an LLM that generates arbitrary text, it chooses between predefined options, answers yes-or-no questions, or rates inputs on an ordered scale. Each decision includes calibrated confidence, making it possible to send uncertain cases to a larger model or a human while handling routine decisions quickly. The first model in the family, strands-decider-2B, has 1.9 billion parameters and is based on a Qwen3.5-2B-Base torso with its language-modeling head replaced by a lightweight pointer head. It evaluates answers in one forward pass without text generation or a decoding loop. Supported decision types include choice, noul for probability-of-true questions, and score for ordered scales. The model is designed for tasks such as model routing, tool selection, tool-argument checking, request triage, grounding and safety guardrails, policy classification, and low-cost evaluation.
  • 18
    Clef-flash

    Clef-flash

    Cloudflare

    Clef-flash is a fast 9B multimodal decision model that turns a state and a schema of typed questions into structured decisions. It can read state information as text, JSON, images, or video and returns a probability for every allowed option of every question. The model supports a 24,576-token context window and vision input, allowing developers to evaluate structured application data and visual content within the same workflow. Clef Flash supports three typed question formats: noul for estimating the probability that a proposition is true, choice for selecting among predefined options with per-option probabilities, and score for evaluating an input against ordered criteria and returning a probability-weighted score. Applications can submit from 1 to 64 questions in a single request, with answers returned under the same question IDs. Example uses include deciding whether a support request is urgent, routing it to the correct team, and scoring the severity of customer impact.
  • 19
    Span-01

    Span-01

    Respan

    Span-01 is a hyper-parallel reasoning classifier built to detect complex behaviors across long production traces in a single forward pass. Rather than generating tokens sequentially, it reasons over natural-language behavior definitions and returns direct probabilities for present, absent, and not_observable. The model is trained for general classification reasoning with RLAIF before being specialized for behavior detection, allowing it to apply previously unseen definitions to new behaviors and domains instead of relying on a fixed taxonomy. At inference time, hyper-parallel definition branches evaluate multiple behaviors against the same trace context simultaneously, while hybrid attention helps preserve reasoning over long, complex inputs. Span-01 is designed for production monitoring across safety, reliability, grounding, agent behavior, and response quality, including jailbreak and prompt injection detection, privacy, and secrets.
  • 20
    Rayline Decisions API
    Rayline Decisions API is a hosted AI decision service for developers building applications and agent workflows. A caller submits an input, such as a prompt, conversation, or document, together with questions about that input. The API returns structured decision signals, including choices, probabilities, and scores, which an application can use to select a model or guide another workflow. Rayline describes the service as trained for model routing and compatible with Jev-style requests. Developers can use these responses to make routing decisions according to their own model options, policies, or scoring criteria. Decisions API is currently presented as a preview and is priced by input-token usage. It is intended for routing developers and teams adding configurable decision capabilities to AI-enabled software.
  • 21
    Tev1

    Tev1

    Together AI

    Tev1 is a fast decision and classification model built by Together AI and fine-tuned from Qwen3.5 4B. It is designed around a simple interface: provide a piece of state, a question, and 2 to 24 labeled options, and the model returns exactly one letter that software can map back to a semantic key. This structure supports production tasks such as routing support tickets, evaluating automated returns, applying policy rules, rating sentiment, categorizing documents, and inserting deterministic decision points into larger agent pipelines. The model has 4.7B parameters, a 32.8K-token context window, and runs with non-thinking generation for fast, predictable classification. It was trained on 37,840 examples spanning intent detection, boolean comprehension, sentiment, policy application, routing, news categories, and research-paper taxonomy. Recommended settings use temperature 0 and a very small output budget so responses remain stable and easy to parse.
    Starting Price: $0.04 per 1M tokens
  • 22
    SubQ Snap

    SubQ Snap

    Subquadratic

    SubQ Snap is a long-context decision model for enterprise and agentic workflows, built to make structured decisions quickly over large amounts of information. It handles tasks such as choosing which files are relevant, selecting the next tool, checking whether a result passes, deciding whether a workflow should continue, and determining when a frontier model is needed. Instead of generating open-ended text, Snap takes context, a question, and a finite set of allowed answers, then returns a probability distribution over those answers. It supports up to 384K tokens of mixed text and image context, allowing it to reason over codebases, contracts, interaction histories, logs, screenshots, charts, diagrams, and scanned pages in a single call. Powered by Subquadratic Sparse Attention, Snap reduces the cost of reading long inputs while a decision head maps contextual representations directly to admissible answers without autoregressive generation.
    Starting Price: $0.05 per million input tokens
  • 23
    Sage2

    Sage2

    Levanto Labs

    Sage2 is a cloud-hosted AI model and API for developers building software agents and automated workflows. It accepts structured task instructions and returns decisions or extracted information in machine-readable formats, allowing applications to delegate steps such as classification, routing, evaluation, and data extraction to a model designed for tool-using agents. Developers can call Sage through its API, inspect model behavior in browser-based demos, and connect it to agent development environments including Cursor and Claude Code. The service offers API usage pricing with a free tier and paid monthly plans; usage charges and included credits vary by plan. Sage is intended for individual developers and software teams that want to add decision-making and extraction capabilities to applications without hosting the model themselves.
  • 24
    LFM3

    LFM3

    Liquid AI

    LFM3 is Liquid AI’s family of compact, open-weight foundation models for text, vision, audio, and multimodal tasks. The release includes text-generation, vision-language, audio, omni, and decision-model variants. Developers can download the model weights, adapt supported models through fine-tuning, and run them locally on edge hardware or access them through Liquid AI’s hosted API. Use cases include conversational assistants, tool calling, speech transcription and generation, image and video understanding, document extraction, classification, routing, and robotics applications. LFM3 is intended for developers, researchers, and organizations building specialized AI applications that need flexible deployment choices, including local inference. Open-weight models are available under Liquid AI’s LFM Open License; commercial eligibility and terms depend on an organization’s annual revenue.
  • 25
    Marengo

    Marengo

    TwelveLabs

    Marengo is a multimodal video foundation model that transforms video, audio, image, and text inputs into unified embeddings, enabling powerful “any-to-any” search, retrieval, classification, and analysis across vast video and multimedia libraries. It integrates visual frames (with spatial and temporal dynamics), audio (speech, ambient sound, music), and textual content (subtitles, overlays, metadata) to create a rich, multidimensional representation of each media item. With this embedding architecture, Marengo supports robust tasks such as search (text-to-video, image-to-video, video-to-audio, etc.), semantic content discovery, anomaly detection, hybrid search, clustering, and similarity-based recommendation. The latest versions introduce multi-vector embeddings, separating representations for appearance, motion, and audio/text features, which significantly improve precision and context awareness, especially for complex or long-form content.
    Starting Price: $0.042 per minute
  • 26
    Colabeler

    Colabeler

    Colabeler

    Image classification, bounding box, polygon, curve, 3D localization Video trace, text classification, text entity labeling. Support custom task plugin, you can create your own label tool. Export PascalVoc XML (The same format used by ImageNet) and CoreNLP file. Supports Windows/Mac/CentOS/Ubuntu.
  • 27
    Alegion

    Alegion

    Alegion

    Alegion is the data labeling solution for enterprise-grade Machine Learning. We lead the industry in streaming, high-resolution, high-density video annotation, delivering accurately-annotated, model-ready data to train and validate ML models. Alegion provides both the platform and workforce to operate with quality at scale, processing structured and unstructured data including video, image, audio, and text. Our ML powered platform speeds up task completion by as much as 70%, including classless object tracking and single click smart polygon generation. Segmentation options include Keypoint, Bounding Box, Polyline, & Polygon segmentation, for image and video. Semantic Segmentation tools deliver seamless entity boundaries with pixel perfect accuracy. NLP and NER capabilities support text and audio classification and sentiment analysis. The platform is highly configurable to support hybrid use cases. Available via SaaS (Alegion Control), Managed Platform, and Managed Labeling Services.
  • 28
    Seed Audio 1.0
    Seed Audio 1.0 is a non-streaming audio generation API based on HTTP, designed to generate complete audio from text prompts, reference audio, or reference images. It supports text-only generation, where audio is created directly from the prompt; reference-audio generation, where uploaded reference clips guide the output; and reference-image generation, where an image reference can be passed to generate audio from the text to be synthesized. Built as part of BytePlus Seed Speech, Audio 1.0 uses the seed-audio-1.0 model version and is positioned as an audio creation capability rather than a standard speech-only endpoint. It can generate voice, music, and sound effects in a single pass, making it useful for producing richer audio scenes without separately creating and mixing every track. The API is intended for developers building audio generation into applications, workflows, and production systems, with a request-based structure that lets teams submit prompts.
  • 29
    Nyckel

    Nyckel

    Nyckel

    Nyckel makes it easy to auto-label images and text using AI. We say ‘easy’ because trying to do classification through complex “we-do-it-all” AI/ML tools is hard. Especially if you’re not a machine learning expert. That’s why Nyckel built a platform that makes image and text classification easy for everyone. In just a few minutes, you can train an AI model to identify attributes of any image or text. Whether you’re sorting through images, moderating text, or needing real-time content labeling, Nyckel lets you build a custom classifier in just 5 minutes. And with our Classification API, you can auto-label at scale. Nyckel’s goal is to make AI-powered classification a practical tool for anyone. Learn more at Nyckel.com.
  • 30
    Azure AI Speech
    Build voice-enabled apps confidently and quickly with the Speech SDK. Transcribe speech to text with high accuracy, produce natural-sounding text-to-speech voices, translate spoken audio, and use speaker recognition during conversations. Create custom models tailored to your app with Speech studio. Get state-of-the-art speech to text, lifelike text to speech, and award-winning speaker recognition. Your data stays yours, your speech input is not logged during processing. Create custom voices, add specific words to your base vocabulary, or build your own models. Run Speech anywhere, in the cloud or at the edge in containers. Quickly and accurately transcribe audio in more than 92 languages and variants. Gain customer insights with call center transcription, improve experiences with voice-enabled assistants, capture key discussions in meetings and more. Use text to speech to create apps and services that speak conversationally, choosing from more than 215 voices, and 60 languages.
  • 31
    UHRS (Universal Human Relevance System)
    When you need transcription, data validation, classification, sentiment analysis, or other related tasks, UHRS can give you what you need. We provide human intelligence to train machine learning models to help you solve some of your most challenging problems. We make it easy for judges to access UHRS anywhere, at any time. All that’s needed is an internet connection, and judges are good to go. Work on tasks like video annotation in just a few minutes. With UHRS, you can classify thousands of images quickly and easily. Train your products and tools with improved image detection, boundary recognition, and more with high quality annotated image data. Classify images, semantic segmentation, object detection. Validating audio to text, conversation, and relevance. Identify sentiment of a tweet, and document classification. Ad hoc data collection tasks, information correction/moderation, and survey.
  • 32
    Oracle Generative AI Service
    Generative AI Service Cloud Infrastructure is a fully managed platform offering powerful large language models for tasks such as generation, summarization, analysis, chat, embedding, and reranking. You can access pretrained foundational models via an intuitive playground, API, or CLI, or fine-tune custom models on your own data using dedicated AI clusters isolated to your tenancy. The service includes content moderation, model controls, dedicated infrastructure, and flexible deployment endpoints. Use cases span industries and workflows; generating text for marketing or sales, building conversational agents, extracting structured data from documents, classification, semantic search, code generation, and much more. The architecture supports “text in, text out” workflows with rich formatting, and spans regions globally under Oracle’s governance- and data-sovereignty-ready cloud.
  • 33
    GPT-5 nano
    GPT-5 nano is OpenAI’s fastest and most affordable version of the GPT-5 family, designed for high-speed text processing tasks like summarization and classification. It supports text and image inputs, generating high-quality text outputs with a large 400,000-token context window and up to 128,000 output tokens. GPT-5 nano offers very fast response times, making it ideal for applications requiring quick turnaround without sacrificing quality. Pricing is extremely competitive, with input tokens costing $0.05 per million and output tokens $0.40 per million, making it accessible for budget-conscious projects. The model supports advanced API features such as streaming, function calling, structured outputs, and fine-tuning. While it supports image input, it does not handle audio input or web search, focusing on core text tasks efficiently.
    Starting Price: $0.05 per 1M tokens
  • 34
    gpt-4o-mini Realtime
    The gpt-4o-mini-realtime-preview model is a compact, lower-cost, realtime variant of GPT-4o designed to power speech and text interactions with low latency. It supports both text and audio inputs and outputs, enabling “speech in, speech out” conversational experiences via a persistent WebSocket or WebRTC connection. Unlike larger GPT-4o models, it currently does not support image or structured output modalities, focusing strictly on real-time voice/text use cases. Developers can open a real-time session via the /realtime/sessions endpoint to obtain an ephemeral key, then stream user audio (or text) and receive responses in real time over the same connection. The model is part of the early preview family (version 2024-12-17), intended primarily for testing and feedback rather than full production loads. Usage is subject to rate limits and may evolve during the preview period. Because it is multimodal in audio/text only, it enables use cases such as conversational voice agents.
  • 35
    SAM Audio
    SAM Audio is a next-generation AI model for detailed audio segmentation and editing. It lets users isolate specific sounds from complex audio mixtures using intuitive prompts that mimic how people think about sound. You can type descriptive text (like “remove dog barking” or “keep vocals only”), click on objects in a video to pull their associated audio, or mark specific time spans where target sounds occur — all in one unified system. SAM Audio is available for experimentation and integration through Meta’s Segment Anything Playground platform, where users can upload their own audio or video files and instantly try SAM Audio’s capabilities. It’s also downloadable for use in custom audio and research workflows. Unlike traditional audio tools that focus on single, narrow tasks, SAM Audio supports multiple kinds of prompts and real-world sound environments with high accuracy.
  • 36
    Azure Speech to Text
    Quickly and accurately transcribe audio to text in more than 85 languages and variants. Customize models to enhance accuracy for domain-specific terminology. Get more value from spoken audio by enabling search or analytics on transcribed text or facilitating action, all in your preferred programming language. Get accurate audio to text transcriptions with state-of-the-art speech recognition. Add specific words to your base vocabulary or build your own speech-to-text models. Run Speech to Text anywhere, in the cloud or at the edge in containers. Access the same robust technology that powers speech recognition across Microsoft products. Convert audio to text from a range of sources, including microphones, audio files, and blob storage. Use speaker diarisation to determine who said what and when. Get readable transcripts with automatic formatting and punctuation. Tailor your speech models to understand organization- and industry-specific terminology.
  • 37
    GPT-5 mini
    GPT-5 mini is a streamlined, faster, and more affordable variant of OpenAI’s GPT-5, optimized for well-defined tasks and precise prompts. It supports text and image inputs and delivers high-quality text outputs with a 400,000-token context window and up to 128,000 output tokens. This model excels at rapid response times, making it suitable for applications requiring fast, accurate language understanding without the full overhead of GPT-5. Pricing is cost-effective, with input tokens at $0.25 per million and output tokens at $2 per million, providing savings over the flagship model. GPT-5 mini supports advanced features like streaming, function calling, structured outputs, and fine-tuning, but does not support audio input or image generation. It integrates well with various API endpoints including chat completions, responses, and embeddings, making it versatile for many AI-powered tasks.
    Starting Price: $0.25 per 1M tokens
  • 38
    T5

    T5

    Google

    With T5, we propose reframing all NLP tasks into a unified text-to-text-format where the input and output are always text strings, in contrast to BERT-style models that can only output either a class label or a span of the input. Our text-to-text framework allows us to use the same model, loss function, and hyperparameters on any NLP task, including machine translation, document summarization, question answering, and classification tasks (e.g., sentiment analysis). We can even apply T5 to regression tasks by training it to predict the string representation of a number instead of the number itself.
  • 39
    Illuminarty

    Illuminarty

    Illuminarty

    Detect AI-generated images, synthetic, tampered images, and deep fakes. We offer various functionalities to help our users understand the generation of images through AI. Check whether or not an image has been AI-generated, from which AI model it has been generated, and which regions of the image have been generated. Find out the probability of AI generation for a given image. Illuminarty combines various computer vision algorithms to provide the likelihood of the image being generated from one of the public AI generation models. Find out the probability of AI generation for a given text. Illuminarty combines various NLP algorithms to provide the likelihood of the text being generated from one of the public AI generation models. We denote the specific passages from the given text that are most likely the result of AI generation. We are working on a web browser extension that lets us use our detectors while we surf on the internet.
  • 40
    BilberryDB

    BilberryDB

    BilberryDB

    BilberryDB is an enterprise-grade vector-database platform designed for building AI applications that handle multimodal data, including images, video, audio, 3D models, tabular data, and text, across one unified system. It supports lightning-fast similarity search and retrieval via embeddings, allows few-shot or no-code workflows to create powerful search/classification capabilities without large labelled datasets, and offers a developer SDK (such as TypeScript) as well as a visual builder for non-technical users. The platform emphasises sub-second query performance at scale, seamless ingestion of diverse data types, and rapid deployment of vector-search-enabled apps (“Deploy as an App”) so organisations can build AI-driven search, recommendation, classification, or content-discovery systems without building infrastructure from scratch.
  • 41
    OCI Data Labeling
    OCI Data Labeling is a service that enables developers and data scientists to build accurately labelled datasets for training AI and machine-learning models. It supports documents (PDF, TIFF), images (JPEG, PNG), and text, allowing users to upload raw data, apply annotations (such as classification labels, object-detection bounding boxes, or key-value pairs), and export the results in line-delimited JSON for seamless integration into model-training workflows. The service offers custom templates for different annotation formats, user interfaces, and public APIs for dataset creation and management, and smooth interoperability with other data and AI services, so annotated data can feed directly into custom vision or language models, as well as Oracle’s AI services. OCI Data Labeling lets users create a dataset, generate records, annotate them, and then use the export snapshot for model development.
    Starting Price: $0.0002 per 1,000 transactions
  • 42
    HumanSignal

    HumanSignal

    HumanSignal

    HumanSignal's Label Studio Enterprise is a comprehensive platform designed for creating high-quality labeled data and evaluating model outputs with human supervision. It supports labeling and evaluating multi-modal data, image, video, audio, text, and time series, all in one place. It offers customizable labeling interfaces with pre-built templates and powerful plugins, allowing users to tailor the UI and workflows to specific use cases. Label Studio Enterprise integrates seamlessly with popular cloud storage providers and ML/AI models, facilitating pre-annotation, AI-assisted labeling, and prediction generation for model evaluation. The Prompts feature enables users to leverage LLMs to swiftly generate accurate predictions, enabling instant labeling of thousands of tasks. It supports various labeling use cases, including text classification, named entity recognition, sentiment analysis, summarization, and image captioning.
  • 43
    EmbeddingGemma 2
    EmbeddingGemma 2 is an open, lightweight multimodal embedding model designed to map text, code, images, video, and audio into a shared embedding space for search, retrieval, classification, routing, and RAG applications. Built on the Gemma 4 architecture and released under the Apache 2.0 license, it has 740 million parameters and is optimized for on-device inference. Its modular design can use as little as 270M parameters for text-only workloads, with optional vision and audio encoders for full multimodal support. Matryoshka Representation Learning lets developers reduce output vectors from 768 dimensions to 512, 256, or 128, lowering storage and memory requirements for local vector databases. The model supports an 8K-token context window and can process up to 5.5 minutes of audio, 29 images, 58 video frames, or interleaved combinations on local hardware.
  • 44
    Classify

    Classify

    Classify

    Classify provides page-level contextual intelligence for advertisers, developers, publishers, and software systems that work with the open web. Its ContentGraph index associates individual web pages with classifications, scores, and source context, helping distinguish pages on the same domain by their specific subject. Advertisers can use Classify's Atlas product to define campaign contexts and identify relevant pages, while developers can use the API to classify URLs and search the index by meaning. The Phoenix developer interface provides example requests and structured classification results for integrating page context into applications and agent workflows. Classify offers URL-classification and semantic-search endpoints, with indexed results returned inline when available. The platform is intended for applications and media workflows that need page-level descriptions of web content.
  • 45
    Vyi

    Vyi

    Vyi

    Vyi is an AI creative workspace for generating and editing images, videos, music, and speech. Users can create images from text prompts, edit with reference images, generate videos from text or images, and produce music or spoken audio. Tools also support AI headshots, product photography, background removal, image upscaling and restoration, lip sync, and audio noise reduction. An asset library organizes uploads, generation history, and drafts. Introductory credits are available after sign-in, and one-time credit packs start at $9 with no automatic renewal. Users can submit support requests through their accounts.
    Starting Price: $9 one-time
  • 46
    DeepInfra

    DeepInfra

    DeepInfra

    DeepInfra is an AI inference cloud that makes it simple to run the latest machine learning models at scale, including LLMs, vision models, embeddings, image generation, video generation, speech, and more. It provides serverless inference through simple APIs, allowing developers to integrate production-ready AI models without managing GPU infrastructure, autoscaling, deployment complexity, or model hosting operations. DeepInfra supports OpenAI-compatible APIs for LLMs and embeddings, making it easier to switch from existing OpenAI-style integrations while accessing a broad catalog of open and commercial models. Its Native API gives access to every model type available on the platform, including image generation, speech recognition, object detection, token classification, fill-mask, image classification, zero-shot image classification, and text classification. DeepInfra is optimized for scalable, low-latency inference and runs models on high-performance GPU infrastructure.
  • 47
    Seedance 1.5 pro
    Seedance 1.5 Pro is a next-generation AI audio-video generation model developed by ByteDance’s Seed research team that produces native, synchronized video and sound in a single unified pass from text prompts and image or visual inputs, eliminating the traditional need to create visuals first and add audio later. It features joint audio-visual generation with highly accurate lip-sync and motion alignment, supporting multilingual audio and spatial sound effects that match the visuals for immersive storytelling and dialogue, and it maintains visual consistency and cinematic motion across multi-shot sequences including camera moves and narrative continuity. Able to generate short clips (typically 4–12 seconds) in up to 1080p quality with expressive motion, stable aesthetics, and optional first- and last-frame control, the model works for both text-to-video and image-to-video workflows so creators can animate static images or build full cinematic sequences with coherent narrative flow.
  • 48
    Ximilar

    Ximilar

    Ximilar

    Ximilar is the first MLaaS platform for training and fine-tuning vision-language models without coding, enabling multimodal AI without in-house research teams. Build and train custom models on your own image and text data, then deploy via a single API click. Chain multiple models into automated workflows using Flows. Key capabilities: — Vision-language model fine-tuning on custom datasets — Image classification, annotation, and object detection — Visual search handling thousands of queries per second — Text-to-image search using natural language queries — Automated tagging and product description generation — OCR and text extraction from images — Fashion AI for apparel tagging and visual search — Defect detection for manufacturing and quality control — Classification, grading, and pricing of collectible items Built on Intel Xeon® with TensorFlow and OpenVINO. Deploy via API or offline. GDPR-compliant, EU servers. 15B+ images processed. Clients in 40+ countries.
  • 49
    Cartesia Ink-Whisper
    Cartesia Ink is a family of real-time streaming speech-to-text (STT) models designed to power fast, natural conversations in voice AI applications, acting as the “voice input” layer that converts spoken language into accurate text instantly. Its flagship model, Ink-Whisper, is specifically engineered for conversational environments, delivering ultra-low latency transcription with a time-to-complete-transcript as fast as 66 milliseconds, enabling fluid, human-like interactions without noticeable delays. Unlike traditional transcription systems built for batch processing, Ink is optimized for live dialogue, handling fragmented, variable-length audio through dynamic chunking, which reduces errors and improves responsiveness during pauses, interruptions, or rapid exchanges.
  • 50
    IBM Datacap
    Streamline the capture, recognition and classification of business documents. IBM® Datacap software is a key capability of the IBM Cloud Pak® for Business Automation. It streamlines the capture, recognition and classification of business documents. Its natural language processing, text analytics and machine learning technologies identify, classify and extract content from unstructured or variable paper documents. Supports multichannel input from scanners, faxes, emails, digital files such as PDF, and images from applications and mobile devices. Uses machine learning to automate the processing of complex or unknown formats and highly variable documents difficult to capture with traditional systems. Enables you to export documents and information to a range of applications and content repositories from IBM and other vendors. Offers configuration of capture workflows and applications using a simple point-and-click interface to speed deployment.