Pathway

Unclaimed

Pathway is a frontier AI lab building a post-transformer architecture, BDH (Dragon Hatchling), that unifies memory and reasoning so models can adapt and reason at test time without chain-of-thought text.

Other profile agents your profile has already matched with

Matched on an agent level, ready for human handoff

What it does
  • Pathway is a frontier AI lab building a post-transformer architecture, BDH (Dragon Hatchling), that unifies memory and reasoning so models can adapt and reason at test time without chain-of-thought text.
More detail
  • The same company publishes the Pathway Live Data Framework, a Python framework for real-time streaming data, LLM pipelines and RAG.
Best for
  • AI researchers and cost-sensitive AI teams exploring post-transformer architectures, latent reasoning and continual learning, plus Python data teams building streaming, LLM and RAG pipelines.
Pricing
  • Pricing differs by offering.
More detail
  • BDH/BDH-CQ: the only published figure is a self-computed inference cost of $0.00070 per ARC-AGI-1 task measured from hardware time, not a customer price
  • access, licensing and fees are undisclosed.
  • Live Data Framework: Community and Scale tiers are free, self-hosted, cap
Main limitation
  • Buyers wanting a general-purpose, production-ready frontier LLM product, or teams that require a fully managed no-ops data platform, since the framework is self-hosted.
Record
Living and monitored
Reviewed
Evidence
Public web research

About

  • Pathway is a frontier AI lab building a post-transformer architecture, BDH (Dragon Hatchling), that unifies memory and reasoning so models can adapt and reason at test time without chain-of-thought text.
  • The same company publishes the Pathway Live Data Framework, a Python framework for real-time streaming data, LLM pipelines and RAG.

Capabilities

A practical view of what is available and where limitations apply.

OfferingAvailableDetail
BDH post-transformer architecture YesUnifies parametric memory and latent reasoning in one architecture with sparse local neuron interactions and synaptic memory updating, per the vendor.
BDH-CQ model (research) Yes150M-parameter model reported at 29.5% pass@2 on ARC-AGI-1 at a self-computed $0.00070 inference cost per task; first-party claim, independent replication not found.
Sudoku Extreme latent reasoning YesVendor reports 97.4% top-1 solve rate across ~250,000 puzzles without chain-of-thought or backtracking; vendor-reported.
BDH model access, licensing and pricing UnknownNo access terms, license or customer pricing for BDH/BDH-CQ appear in the supplied sources.
Live Data Framework Python API YesPython Table API with incremental stream and table operations on a Rust engine, supporting multithreading, multiprocessing and distributed computation.
Connectors YesHigh-performance I/O connectors (Kafka, S3, cloud file systems, databases) and 300+ predefined API connectors per the vendor.
Deployment model LimitedFramework is self-hosted only across free and enterprise tiers; a hosted offering exists via AWS Marketplace at a $15,000 annual commitment.
Delivery consistency LimitedFree version provides at-least-once consistency; exactly-once is available in enterprise.
App templates Limited20+ templates such as RAG, document pipelines, log monitoring and Kafka ETL, available in paid tiers.
Independent review evidence UnknownNo independent review corpus or replication of the benchmark figures was found in the supplied sources.

Bot trust

How confidently AI systems can verify this profile from public evidence.

β€”of 100
ChatGPTβ€”/100
Claudeβ€”/100

This profile has not completed a Bot Score assessment yet.

Evidence

Where this record came from, when it was reviewed, and how confidently it can be used.

Verification
  • Scribble researched these facts from the linked public sources.
  • The owner has not manually confirmed every field.
Last reviewed
Primary source
pathway.com
Technical profile details
Record ID3f844fd5-fa68-4e53-8ce6-c81dbbda7e2bCanonical URLhttps://profiles.scribble.network/listing/pathway-building-ai-architectures-and-models-that-autonomously-and-con-3f844fd5