January 2026 Summaries
2 posts from OpenRouter
Filter
Month:
Year:
Post Summaries
Back to Blog
OpenRouter's January release introduces several enhancements, including partition sorting that allows users to prioritize faster providers by setting throughput or latency thresholds, ensuring efficient performance without added latency. The Provider Explorer feature enables users to view all providers in one location, highlighting DeepInfra's extensive model offerings and OpenAI's proprietary models. Users can now report bugs and feedback directly from various sections of the platform to aid in assessing provider performance. The Auto Router feature has been upgraded to support 58 models, including Opus 4.5, with customizable model selection using wildcard syntax and no additional markup costs. The SDK Skills Loader allows for the seamless integration of specific skills into any model's context, ensuring idempotency. Additionally, the LLM Leaderboard and gateway performance have seen significant speed improvements, with the leaderboard loading over 50% faster and the gateway achieving a 70% improvement in latency, making it the fastest in the company's benchmarks.
Jan 09, 2026
329 words in the original blog post.
OpenRouter's January Release Spotlight highlights several significant updates, including enhanced features for Partition Sorting, Provider Explorer, Bug & Feedback Reporting, Auto Router customization, and the SDK Skills Loader. The Partition Sorting feature allows users to prioritize fast providers by setting throughput or latency thresholds, while the Provider Explorer consolidates all providers into one accessible location, noting DeepInfra's extensive model collection and OpenAI's proprietary models. Users can now report bugs and feedback more efficiently through various platforms, aiding in assessing provider performance. The Auto Router has expanded its support to 58 models with new customization options, and the Skills Loader in the OpenRouter SDK enables the automatic injection of domain-specific skills with idempotency. Additionally, the LLM Leaderboard has been optimized to be over 50% faster by improving lazy loading techniques, and the gateway now boasts a 70% increase in speed, marking significant latency improvements.
Jan 09, 2026
301 words in the original blog post.