name,owner,accessRoute,source,accessUrl,strength,limitation,evidenceStatus,checkedAt,inferenceTested
Jev,TypeSafe AI,Hosted API,https://typesafe.ai/,https://docs.typesafe.ai/,Typed semantic judgments and confidence estimates.,Closed service; calibrate thresholds on your own traffic.,Primary source checked,2026-11-15,False
Microsoft-Decision-1,Microsoft,Microsoft Foundry / OpenRouter,https://commandline.microsoft.com/microsoft-decision-1-model-foundry/,https://ai.azure.com/catalog/models/Microsoft-Decision-1,"Routing, classification and verification through structured scoring.",Vendor benchmark leadership is scoped to its own evaluation; verify deployment region.,Primary source checked,2026-11-15,False
Gemini Decision 1,Google DeepMind,Google AI Studio / Gemini API,https://blog.google/innovation-and-ai/technology/developers-tools/gemini-decisions-api/,https://aistudio.google.com/,"Decisions over text, images, audio, video and PDFs.",Confirm account quotas and regional data handling before use.,Primary source checked,2026-11-15,False
Decider v1.3 / Decider Flash v1.3,Perplexity,Downloadable weights / vendor API,https://huggingface.co/perplexity-ai/pplx-decider-v1.3,https://huggingface.co/perplexity-ai,"Model cards, typed probabilities and deployment control.",Separate and isolated question modes can differ; validate batching semantics.,Primary source checked,2026-11-15,False
Decisions API,OpenAI,OpenAI API,https://developers.openai.com/api/docs/guides/decisions,https://developers.openai.com/api/docs/guides/decisions,"Predicate, choice and score questions through a dedicated endpoint.",Its input and question schema differs from Jev; handle refusals explicitly.,Primary source checked,2026-11-15,False
Clef / Clef-flash / Clef-omni,Cloudflare,Workers AI / downloadable weights,https://blog.cloudflare.com/clef-decision-models/,https://developers.cloudflare.com/workers-ai/,Managed edge access and open deployment options.,"Prices, modalities and license conditions belong to each exact checkpoint.",Primary source checked,2026-11-15,False
d1 family,Liquid AI,Liquid console / selected downloadable checkpoints,https://www.liquid.ai/blog/d1-ultra,https://console.liquid.ai/,Text and image judgments; hosted and local options across the family.,Hosted d1-ultra and open d1 checkpoints have different capabilities and terms.,Primary source checked,2026-11-15,False
Mammoth-1 / Elephant-1 / Joey-1,Base Labs / Baseten,Baseten API / dedicated deployment / weights,https://www.baseten.co/library/mammoth-1/,https://www.baseten.co/library/mammoth-1/,Mammoth offers text/image decisions and an MIT license.,Dedicated GPU capacity and cost need workload measurements; sibling details require their own cards.,Primary source checked,2026-11-15,False
GLiDE v1.2 / GLiNER2.5-Decide,Fastino Labs,Fastino API / downloadable weights,https://fastino.ai/blog/training-glide-sota-decision-model,https://fastino.ai/,GLiDE v1.2 distills reasoning into direct decision predictions.,Original adaptive-thinking GLiDE and v1.2 are different releases; check the Community License.,Primary source checked,2026-11-15,False
Drex Max v1.1 / Drex 2.0,Nace.AI,Nace console / API,https://www.nace.ai/drex/decision-index,https://console.nace.ai/,Vendor publishes a dated view of decision quality and task costs.,The comparison is hosted by a competing vendor; it is not an independent PALO measurement.,Primary source checked,2026-11-15,False
SubQ Snap 1.2,Subquadratic,SubQ console / Decisions API,https://docs.subq.ai/using-the-api/models/,https://console.subq.ai/,Versioned decision endpoint for text and images.,Use a versioned ID; latest aliases move. Confirm long-context behavior on your workload.,Primary source checked,2026-11-15,False
Strands Decider 2B,Strands Labs,Repository / self-hosted server,https://github.com/strands-labs/strands-decider,https://github.com/strands-labs/strands-decider,Small reference decision model with training and serving material.,A Strands Labs repository does not establish an AWS managed production service.,Primary source checked,2026-11-15,False
CLM-8B,Contrastive-LM research project,Repository / self-hosted weights,https://github.com/Contrastive-LM/CLM,https://github.com/Contrastive-LM/CLM,Contrastive matching between states and candidate actions.,Action caching and benchmark speedups depend on the evaluated task and hardware.,Primary source checked,2026-11-15,False
D3 Turbo / d3-flash / d3-mini,vLLM Semantic Router,Model cards / self-hosted serving,https://huggingface.co/vllm-sr/d3-turbo,https://huggingface.co/vllm-sr,Multimodal decision family with an Apache-2.0 D3 Turbo card.,Check each checkpoint and serving implementation; a demo API is not a service guarantee.,Primary source checked,2026-11-15,False
Laya / laya-mlx,ConvAI Innovations,Repository / local execution,https://github.com/NandhaKishorM/laya,https://github.com/NandhaKishorM/laya,Local non-autoregressive typed text decisions.,Compact models and hardware-specific timings need domain and device validation.,Primary source checked,2026-11-15,False
Sage / Sage2,Levanto Labs,Levanto platform / API and streaming,https://docs.levanto.ai/,https://platform.levanto.ai/,"Typed decisions, extraction and streaming updates with explicit null answers.",Current documentation uses Sage; confirm the exact Sage2 model identifier and version before integration.,Primary source checked,2026-11-15,False
Mercury Decide,Inception,Inception API; service wind-down announced,https://docs.inceptionlabs.ai/,https://docs.inceptionlabs.ai/,"Typed classification, routing, scoring and span extraction.",Official documentation announces API closure after 28 December 2026; reassess new integrations.,Primary source checked,2026-11-15,False
Celeris-1.1 Decision,Celeris,System One / Decisions API,https://docs.celeris.ai/decisions,https://docs.celeris.ai/decisions,Documented text/image decisions and two API formats.,No streaming; confirm pricing and limits for the selected API route.,Primary source checked,2026-11-15,False
Bivia,Articul8,Articul8 platform / API,https://articul8.ai/bivia,https://app.articul8.ai/,Vendor documents multimodal typed decisions and an abstention option.,Validate each claimed modality and operation mode; the product page is not independent qualification.,Primary source checked,2026-11-15,False
Quyet-1.0-Large,Independent research,Hugging Face weights / quyet package,https://huggingface.co/chinhnc/Quyet-1.0-Large,https://huggingface.co/chinhnc/Quyet-1.0-Large,A documented local Gemma-based decision model with typed questions.,"The card lists 62.5 GB bf16 weights; check GPU capacity and both LICENSE and NOTICE, including base-model terms.",Primary source checked,2026-11-15,False
H2O-Lightning-4B v1.1,H2O,H2O product page / open weights / Enterprise h2oGPTe,https://h2oai.com/products/lightning,https://h2oai.com/products/lightning,Open-weight decision family for text and up to four images per record.,"The product page covers the family; confirm the exact 4B v1.1 checkpoint, serving adapter and licensing terms.",Primary source checked,2026-11-15,False