what's OpenRouter?
OpenRouter is a unified API platform that provides access to hundreds of AI language models from various providers, enabling developers to route requests based on factors like cost, speed, and reliability. It supports features such as bring-your-own-key integration, real-time throughput tracking, and advanced capabilities like streaming reasoning summaries and tool calling. The platform simplifies AI development by normalizing APIs across providers, including support for OpenAI-compatible endpoints and custom routing preferences for privacy or performance. In addition to its AI focus, OpenRouter incorporates cryptocurrency functionalities, allowing users to purchase credits via crypto payments through integrations like Coinbase Commerce and on-chain transactions. This enables automated, headless funding for AI usage, with options for generating invoices and supporting multiple chains. The company has conducted funding rounds and a token sale, as documented on platforms like CryptoRank, positioning it at the intersection of AI infrastructure and blockchain payments.
OpenRouter launched SDK Skills Loader for model skills and a bug/feedback feature for provider degradation.
Links
x.com/openrouteraiReleased SDK Skills Loader feature that enables loading and using skills in any model's context.
Launched a bug reporting and feedback feature that allows users to report issues or provide feedback about any generation, with the stated purpose of quantifying provider degradation.
Added MiniMax M2.1 model to the platform on December 23, 2025, featuring improvements in coding tasks and token efficiency compared to M2.
Formed a partnership or collaboration with IDLHUB.
The platform launches a new Broadcast feature that sends all traces directly to LangSmith without code changes, working across LangChain, provider SDKs, and the OpenRouter SDK.
The platform launches integration with Langfuse, enabling users to send project traces to self-hosted Langfuse instances for LLM observability.
Native Claude Code support launches, enabling its use across all models on the platform.
An empirical study processes 100 trillion tokens across hundreds of models via the platform.
The platform adds a debugging capability for developers to echo raw upstream requests.
The platform processed over 1 trillion tokens daily throughout last week, compared to OpenAI's API volume of approximately 8.6T tokens per day in October.
INTELLECT-3 model launches on the platform with full open-source release including weights, code, environments, and detailed documentation.
FLUX.2 image models launch on the platform with two variants: 'pro' for frontier-level quality with fast, low-cost generation, and 'flex' optimized for complex text and typography.
Grok 4.1 Fast API launches on the platform with 2M token tool-calling capabilities, priced at $0.20/1M input and $0.50/1M output tokens, with free access available until December 3.
Gemini 3 is now live on the platform.
LOGIC, a new verification method for LLM inference in trustless environments, works out of the box with the platform to detect model substitution, quantization, and decode-time attacks with low computational overhead.
OpenRouter announces immediate hiring for a new forward-deployed engineer role to solve complex AI inference issues for a wide variety of companies.
Used in Karpathy's nanochat project; new API parameter for synthetic data generation coming soon.
Chutes achieves #1 ranking for open source inference with over 40 billion tokens processed daily, surpassing competitors DeepInfra, Bedrock and Fireworks. OpenRouter red flagged Chutes for lacking full TEE security, with implementation coming soon.
Bittensor ranks number one on the platform for Subnet 64.