Real-Time Swap Pricing in 1.5 ms — Without the Expensive Infrastructure
RFQ latency is a deal killer

Lean architecture and purpose-built language help banks catch up in the RFQ race.
The Problem:
The RFQ Latency - a Deal Killer
In electronic fixed income markets, the shift to request-for-quote (RFQ) workflows has been swift, but uneven. Top-tier investment banks have optimised their stacks for real-time pricing with latencies in the sub-millisecond range.
But many non-tier one banks and dealers are still catching up, either by experimenting with RFQs via platforms like Bloomberg and Tradeweb, or relying on manual quoting with slower, legacy systems.
The challenge:
The Need For Speed
When RFQs expire in milliseconds, a slow system isn’t just inefficient, it’s invisible.
The Solution:
Smart Design Over Heavy Tech
At Algorithmica, we’ve built a lean, real-time pricing system that competes with the best. Without GPU farms, parallelism, or high-end infrastructure.
Our Speed Test Aims:
- Full support for discounts and multiple forwarding curves
- Real-time quote handling from Bloomberg (IRS, FRAs, OIS, etc.)
- Swap pricing across a typical dealer portfolio
- Total latency target: sub-2 ms on a standard PC
Target beaten.
The Test Case
Here’s what we ran, continuously in production-like conditions:
Step | Description | Time
1 Bloomberg Feed | 50 instruments updating in real-time | ~0.4 ms
2 Curve Calibration | Discount curve + multiple forward curves | ~0.6 ms
3 Swap Revaluation | 10 swaps, 10Y maturity each | ~0.5 ms
Total Time | End-to-end pricing | ~1.5 ms
This is single-threaded, no GPU, no batching.
Clean code, lean architecture and Qlang, our proprietary domain-specific language that compiles to fast machine code using the LLVM toolchain.
Why Qlang?
We built Qlang to serve the needs of pricing and risk:
- First-class support for vectors, curves, and market conventions
- Runtime compilation to LLVM-generated machine code
- Seamless integration with our pricing servers
This allows us to express pricing logic at a high level, but run it as fast as hand-optimised C++, with none of the boilerplate.
What this means for non-tier one banks and dealers
We’ve proven that you don’t need a big bank budget to compete in electronic fixed income trading.
You can:
- Respond to RFQs fast enough to win
- Maintain accurate, curve-driven pricing logic
- Use low-cost hardware and simple deployments
- Scale incrementally as your volumes grow
A level playing field in RFQs
If you’re a bank or dealer moving towards RFQs on Bloomberg or Tradeweb, or upgrading your pricing infrastructure, let’s talk.