AI research company Pathway has secured additional funding at a $500 million valuation, bringing its total seed financing to $30 million as it prepares to scale a new generation of models built around its post-Transformer architecture.
The financing includes participation from Id4 Ventures, TQ Ventures, Red Bridge Ventures, Kadmos Capital, and WS Investment Co., the investment arm of Wilson Sonsini, alongside Databricks Chief AI Scientist Jonathan Frankle. Pathway said much of the new capital will go toward expanding compute capacity, including NVIDIA GB300 systems.
The investment arrives alongside new results for Pathway’s BDH-CQ reasoning model, giving the funding a more concrete technical backdrop than the increasingly common promise of simply building larger foundation models. Pathway says its 150-million-parameter model scored 29.5% on the public ARC-AGI-1 evaluation set at a calculated inference cost of just $0.0007 per task.
Funding Arrives as Pathway Bets Against Brute-Force Scaling
Much of the current frontier-model race has centered on scale: larger clusters, more training data, more parameters, and increasing amounts of test-time compute. Pathway is taking a different approach, arguing that some of the industry’s cost and memory limitations are architectural rather than problems that can indefinitely be solved by adding more compute.
That thesis is at the center of its BDH, or Dragon Hatchling, architecture. Pathway describes BDH as a post-Transformer design intended to combine reasoning, memory, and adaptation within the model itself rather than relying primarily on an expanding context window or external memory systems.
The architecture draws on ideas including persistent state, sparse activity, local interactions, and continual adjustment. Pathway describes the design as biologically inspired rather than an attempt to directly reproduce how the brain works.
The new financing gives the company additional resources to test whether those characteristics continue to hold as BDH models become substantially larger.
Pathway is also adding Adam Kurzrok as Chief Product Officer. Kurzrok previously served as a Group Product Manager for Gemini at Google DeepMind and will lead product direction around packaging, evaluating, and deploying BDH-based models.
The company is also formalizing an advisory group that includes Transformer co-inventor Łukasz Kaiser, Frankle, NYU professor Martín Farach-Colton, and economist Jacques Attali.
What Makes BDH-CQ Different
BDH-CQ is an extension of the broader BDH architecture designed around in-context learning and recurrent latent reasoning.
Instead of requiring a model to express much of its intermediate reasoning as generated text, BDH-CQ performs iterative computation inside a continuous latent workspace. Examples supplied during inference update a recurrent memory state, after which the system works through the new problem internally before decoding its answer. The model’s parameters remain fixed during this process.
This matters because token-based reasoning can become expensive as reasoning traces lengthen. In conventional chain-of-thought approaches, intermediate steps are generated sequentially and then fed back into later stages of the reasoning process.
BDH-CQ is designed to keep more of that computation within its internal state rather than continuously translating intermediate reasoning into language.
The distinction does not mean language-based reasoning is inherently obsolete. Rather, Pathway is testing whether some reasoning workloads can be handled more efficiently without requiring every intermediate computational step to be serialized as text.
ARC-AGI Results Put Cost Efficiency at the Center
The most notable result accompanying the financing comes from ARC-AGI-1, a benchmark designed to test whether AI systems can infer transformations from a small number of examples and apply those rules to unseen inputs.
Pathway’s 150-million-parameter BDH-CQ achieved 29.5% pass@2 across the 400-task public ARC-AGI-1 evaluation set.
The technical evaluation estimates that each task required roughly 0.85 seconds of NVIDIA H200 GPU time, corresponding to a calculated cost of approximately $0.00070 per task when assuming an H200 cost of $3 per GPU-hour.
Pathway argues that this places BDH-CQ beyond the previously reported cost-versus-accuracy Pareto frontier for ARC-AGI-1.
The significance is not that the 150-million-parameter system now has the highest absolute ARC score. Some larger reasoning systems score higher. Instead, Pathway is focusing on how much reasoning performance the model can deliver for each dollar of inference spending.
That could become increasingly important as enterprises move AI from occasional chatbot interactions toward agents and other systems that may perform large numbers of reasoning operations continuously.
The Technical Results Also Show Where BDH-CQ Struggles
The broader evaluation presents a more nuanced picture than the headline benchmark.
On ConceptARC, which separates reasoning problems into different conceptual categories, BDH-CQ’s performance varied considerably. It performed particularly well on several tasks involving boundary extension and distinguishing filled from unfilled regions, while areas involving copying and ordering proved more difficult.
Controlled experiments produced a similar pattern.
Boundary propagation and copying remained accurate as those operations were extended within the tested ranges. Performance deteriorated more significantly as ordering sequences became longer. Nested relational problems also became more challenging once the required nesting depth increased.
The researchers also found that providing demonstrations closer to the complexity of the target problem could substantially improve performance in some cases, suggesting that part of the limitation concerns how far the system can extrapolate from the examples it receives.
Those weaknesses matter because ARC-AGI remains a specialized visual reasoning benchmark. Strong cost efficiency on ARC does not demonstrate that the same architecture can outperform general-purpose language models across production workloads.
Instead, the results provide evidence for a narrower but potentially consequential idea: sophisticated reasoning capabilities may not always require massive parameter counts or long sequences of generated reasoning tokens.
Pathway Is Also Building Infrastructure for Real-Time AI
Pathway’s work predates its push into post-Transformer models. Alongside BDH, the company develops a data processing framework aimed at streaming data, real-time analytics, large language model applications, and retrieval-augmented generation.
The platform is designed to help AI systems work with information that changes continuously rather than relying entirely on static datasets or periodically rebuilt indexes.
That infrastructure background fits closely with Pathway’s broader research direction. Its existing technology focuses on keeping AI applications synchronized with changing external data, while BDH explores whether persistent memory and adaptation can increasingly become properties of the model architecture itself.
This could matter for AI agents and other long-running systems that need to retain state while reacting to new information over extended periods.
What Comes Next for Pathway
Pathway intends to use the new capital to increase model capacity and train more broadly capable BDH-based systems.
Its roadmap includes mathematical reasoning, ARC-AGI-2 and ARC-AGI-3, as well as development of a large language model incorporating latent reasoning.
The company says early pretraining experiments at scales ranging from 1 billion to 600 billion parameters have shown Transformer-like scaling behavior while retaining characteristics associated with BDH’s latent reasoning approach.
That remains an early indication rather than proof that the architecture’s advantages will persist across large, general-purpose models.
Pathway is ultimately targeting areas such as financial services, healthcare, and technology, where persistent memory, changing information, and inference costs could become particularly important as AI systems move toward longer-running workflows.
For now, the $500 million valuation represents a significant bet on an alternative to the industry’s dominant scaling strategy. The more important test will come as Pathway moves BDH beyond relatively constrained reasoning benchmarks and demonstrates whether its architectural advantages persist across language, mathematics, agents, and real-world applications.
