OpenRouter
@OpenRouter
The Pareto Router is now processing almost 1B tokens per day: https://t.co/IHsAo9CuqH The Auto Router is processing 12B: https://t.co/MewkWfiOm0 See the @theinformation's article below 👇
5retweets54likes
View on XOpenRouter Reaches 13B Daily Tokens as Automated Model Routing Scales
OpenRouter announced its automated routing infrastructure reached 13 billion tokens in daily volume. The Pareto Code router, which selects efficient coding models based on benchmarks, now handles 1 billion tokens daily. It uses a tiered shortlist to route requests to models like
DeepSeek V4 Pro based on user-defined quality thresholds.- Auto Router Volume
- 12B tokens per day
- Pareto Router Volume
- 1B tokens per day
- Pareto Context Window
- 2,000,000 tokens
- Auto Router Customization
- 0-10 cost/quality scale
- Top Pareto Model
- DeepSeek V4 Pro (73.8% share)
This surge follows a Series B funding round and reflects a shift toward multi-model production. By using a meta-model to manage inference (running a trained model to generate outputs), developers hedge against downtime and capture price drops. This layer commoditizes individual models in favor of consistent performance and cost efficiency.
Users can now apply a cost-quality slider to the Auto Router to fine-tune selection. Adjusting a 0–10 scale allows you to prioritize high-intelligence models for reasoning or cheaper alternatives for routine tasks. These settings are available in the routing dashboard, where you can also set guardrails and usage limits.
Still wondering? A few quick answers below.
The Pareto Router is a specialized model selection engine designed for agentic coding tasks. It maintains a curated shortlist of high-performing coding models, such as DeepSeek V4 Pro and GPT-5.4 Mini, ranked by Artificial Analysis benchmarks. Users can set a minimum quality score to automatically route requests to the most cost-effective model that meets their requirements.
The cost-quality slider is a parameter that allows developers to balance model intelligence against token expenses on a scale of 0 to 10. A higher setting directs the Auto Router to prioritize frontier models for complex reasoning, while a lower setting favors more affordable models for routine tasks, helping teams manage their AI inference budgets.
The Pareto Router currently routes traffic to a tiered selection of strong coding models. According to recent usage data, the primary models include DeepSeek V4 Pro, DeepSeek V4 Flash, Kimi K2.6, GPT-5.4 Mini, and Gemini 3.1 Pro. The specific model selected for a request depends on the user's defined quality threshold and the router's performance rankings.
Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →
