Tradeweb Ai-Price: Evolving Credit Pricing for Today’s Market

The electronification of the U.S. credit market over the last ten years has led to huge changes in how clients and dealers trade bonds. From the growth of portfolio trading to parameter-based automated execution to interdealer sweeps, electronification has changed the market microstructure significantly, allowing for greater efficiency and cost savings.

Alongside this electronification, credit pricing algorithms have also seen significant growth across the buy-side and sell-side, with some dealers now systematically quoting trades up to $10 million. Tradeweb Ai-Price has evolved with these changes in the credit market. Whether that is using our pre-trade cost estimates, evaluating dealer responses or assessing execution quality post-trade, Tradeweb Ai-Price supports client decision-making throughout the trade lifecycle.

As reliance on these algorithms continues to increase, we are leveraging advanced machine learning techniques and proprietary datasets to continue to improve Tradeweb Ai-Price. As part of this evolution, we plan to publish a series of content pieces exploring accuracy, methodology enhancements and use cases, with a more detailed whitepaper expected to follow.

Methodology & accuracy

Pricing corporate bonds is a uniquely complex problem as there are many non-linear relationships between bonds in the same issuer, sector and rating buckets. Tradeweb Ai-Price attempts to solve this problem by leveraging a mixture of Gaussian methodology to model the 'change since previous trade' on each side of the market (bid and offer). We use neural network architectures for IG and HY to predict the mean and standard deviation of this 'change', allowing us to capture different market buy/sell dynamics while maintaining an accurate mid estimate. We expect to go into further detail around the methodology behind Tradeweb Ai-Price when we release our full whitepaper.

The charts below show the difference between Tradeweb Ai-Price and the next TRACE print, bucketed by liquidity score as a measure of model accuracy. We considered prints greater than 50k from January 2025 to June 2026 for this analysis and compare our model bids/offers to dealer-to-client trades. The results show around a 30% improvement across IG and HY relative to the benchmark of using the previous TRACE print as a predictor, based on a weighted average across liquidity scores, with each score weighted by trade count. The results demonstrate the model's accuracy across a range of liquidity conditions.

Conclusion

We believe Tradeweb Ai-Price is a state-of-the-art pricing algorithm for U.S. credit markets, leveraging the latest machine learning techniques and both public and proprietary data to support decision-making across the trade lifecycle. Built for the way credit markets trade today, Tradeweb Ai-Price is designed to adapt to changing market conditions and power analytics across our trading workflows and broader platform.

Stay tuned for more content around Tradeweb Ai-Price and the market intelligence it can provide.