Tag
This paper quantifies and expands the theoretical capacity of late-interaction retrieval models, showing that MaxSim can replicate inner products between non-negative vectors and proposing Signed MaxSim for arbitrary real-valued vectors, revealing a representation gap between inner product and late-interaction models.
Anthropic confirms that access to Fable 5 via subscription ends on July 7, after which it will move to usage credits, but plans to restore it as a standard subscription model when capacity improves.
Fable will be removed from subscriptions after July 7th, but the team plans to restore it as soon as capacity allows.
Cerebras' deal with OpenAI to supply $20 billion worth of chips has pre-allocated near-term inference capacity, causing their API waitlist to become effectively infinite for other startups seeking fast inference.
SemiAnalysis refutes the widely circulated claim that half of 2026 US datacenter capacity is delayed or canceled, attributing it to inaccurate models and showing their own data indicates only minor changes in forecasts.
A Stanford, MIT, Harvard, and Anthropic paper explains that larger AI models learn rare skills better because they forget them less during training; their extra capacity protects weak learning signals from being overwritten by common tasks.
This paper investigates why larger models outperform smaller ones, attributing it to reduced gradient interference and better resource allocation, allowing them to learn rare and complex tasks even with infinite data. Experiments on synthetic data and OLMo models verify that larger models avoid overwriting rare-task features due to weaker gradient updates for common tasks.
Sam Altman announces that a program offering compute capacity will be available until the current allocation sells out, with plans to resume later while reserving capacity for ChatGPT and Codex.
Sam Altman announces OpenAI's Guaranteed Capacity, offering discounted tokens for 1-3 year commitments to provide customers with capacity certainty.