December 2024 Summaries

OpenRouter has introduced a new web search feature allowing users to access up-to-date information using any language model on their chatroom, with plans to provide API access in the future. Currently offered for free, this feature aims to enhance information retrieval across various models. Additionally, OpenRouter has implemented price cuts on several language models, including significant reductions for models like nousresearch's Hermes-3 and Meta-Llama's Llama-3.3. Furthermore, a new API endpoint is available in beta, enabling users to access model details and endpoints, although it is subject to changes before its official documentation release.

OpenRouter has introduced new features and updates, including a Web Search functionality that allows users to search for any language model on the OpenRouter Chatroom, making it easier to obtain up-to-date information across various language models. This feature is currently free, with API access expected later. Additionally, there have been price cuts on several models, with reductions ranging from 11% to 31%, impacting models from Qwen, NousResearch, Meta-LLaMA, and Nvidia. A new beta version of the Endpoints API has also been launched, providing users with the ability to access model details and available endpoints, though it is currently undocumented and subject to change.

OpenRouter has introduced the first API designed for scripting headless, on-chain payments specifically catered to large language models (LLMs). This innovative API aims to enable the creation of intelligent agents capable of autonomously funding their operations. The introduction is accompanied by a detailed tutorial available on X, with further documentation provided for those interested in exploring the technology.

OpenRouter has introduced a new feature called Bring Your Own API Keys (BYOK), allowing users to integrate their existing API keys and credits from major providers like OpenAI, Google Cloud, and AWS to enhance their use of OpenRouter. This feature offers several advantages, including higher rate limits by combining provider limits with those of OpenRouter, enabling the use of free models, and utilizing third-party credits efficiently. Additionally, BYOK provides unified analytics by allowing users to track all their large language model (LLM) usage through a single Activity dashboard. The service is available at 5% of the upstream provider's cost, offering the combined benefits of OpenRouter's capabilities and users' pre-existing resources.

Bring Your Own API Keys (BYOK) is a new feature from OpenRouter that allows users to utilize their existing API keys and credits from major providers like OpenAI, Google Cloud, and AWS, thereby increasing their throughput by combining their provider limits with OpenRouter’s. This feature, now available for general access, also enables users to make the most of their third-party credits and offers a unified analytics platform to track all their LLM usage through the OpenRouter Activity dashboard. For a cost of 5% of the upstream provider's fees, users can enjoy the benefits of OpenRouter while leveraging their existing resources.

OpenRouter has introduced a Crypto Payments API, which provides a novel method for scripting headless, on-chain payments for any language model. This innovation allows for the creation of agents capable of autonomously funding their own operations, enhancing the potential for self-sustained artificial intelligence. A tutorial is available on X, guiding users on how to implement this technology, with further information accessible through official documentation.

OpenRouter has announced updates that include support for structured outputs with OpenAI 4o and Fireworks models, allowing for JSON Schema validation and ensuring consistent, type-safe responses, with more providers expected to join soon. Additionally, Google has introduced Gemini Flash 2.0, a new AI offering that promises significantly faster token response times, improved multimodal understanding, enhanced coding capabilities, and advanced function calling. This release will initially have heavy rate limitations to allow shared access among users, and both updates can be explored further through provided links for more detailed information.

OpenRouter has introduced updates that include support for structured outputs and the free availability of Google’s Gemini Flash 2.0. The platform now supports structured outputs for OpenAI 4o and Fireworks models, ensuring JSON Schema validation and type-safe, consistent responses, with plans to add more providers soon. Gemini Flash 2.0, characterized by its significantly faster response time, improved multimodal understanding, enhanced coding capabilities, and advanced function calling, is initially available for free but will be heavily rate-limited to manage user access. Full details on structured outputs can be accessed at OpenRouter's documentation page, and users are encouraged to try the newly released Gemini Flash 2.0 through a dedicated link.

DeepInfra has announced significant price reductions on various models, making them more accessible to users. Notable discounts include a 40% price cut for the Llama 3.2 3B Instruct model, now priced at $0.018, and a substantial 70% reduction for the Mistral Nemo, now available at $0.04. Other models, such as the Qwen2.5 72B Instruct and Mistral 7B Instruct v0.3, have also seen their prices slashed by up to 50%. In addition to these price drops, the Llama 3.3 70b model has been launched with two providers already offering it, expanding the options available to users seeking advanced AI solutions.

DeepInfra has introduced significant price reductions on various models, making advanced AI technologies more accessible. Notable discounts include the Llama 3.2 3B Instruct, which is now 40% cheaper, and the Qwen2.5 72B Instruct, which has seen a 40% price reduction as well. Other models such as the Mistral 7B Instruct v0.3 and Gemma 2 9B have also benefited from substantial price cuts, with reductions of 50% each. Additionally, the newly launched Llama 3.3 70b is already being offered by two providers, highlighting rapid adoption and availability in the market.

OpenRouter has launched a new feature called Author Pages, which allows users to explore models from specific creators by visiting openrouter.ai/<author>, providing detailed statistics for each author's collection and an overview of all their models, along with a carousel feature to discover related models. Additionally, Amazon has introduced its new Nova family of models, which includes Nova Pro 1.0, Nova Micro 1.0, and Nova Lite 1.0, expanding the range of options available for users seeking advanced model capabilities.

OpenRouter has introduced a new feature that allows users to explore models by specific authors, providing detailed statistics for each collection and offering a comprehensive view of all models by an author on a single page. This enhancement includes a carousel feature to discover related models, making it easier to navigate and explore the platform's offerings. Additionally, Amazon has launched a new family of models called Nova, which includes Nova Pro 1.0, Nova Micro 1.0, and Nova Lite 1.0, expanding their lineup with different versions to cater to various needs.