Blogs | Juan Cruz-Benito

Qiskit QuantumKatas: A Benchmark for Evaluating LLMs on Quantum Code

Wed, 27 May 2026 00:00:00 +0000

Together with , I’ve just put a new preprint on arXiv: Qiskit QuantumKatas, a benchmark for measuring how well large language models write quantum computing code.

The starting point was — a well-established, self-paced curriculum for learning quantum computing through programming exercises. The catch is that the originals are written in Q#. We translated them to Qiskit, the most widely-used quantum framework, and packaged the result with an evaluation pipeline built for systematic LLM assessment.

What’s in it

The benchmark has 350 tasks across 26 categories, spanning the full pedagogical arc the QuantumKatas were designed around: fundamental gates and superposition, through canonical algorithms like Deutsch–Jozsa, Simon’s, and Grover’s, up to error correction, key distribution, and quantum games. Each task ships with a natural-language prompt, a canonical solution, and a deterministic test that verifies correctness via classical circuit simulation — so a solution either reproduces the expected statevector or it doesn’t. No approximate matching, no ambiguity.

Building on the QuantumKatas’ existing structure rather than inventing tasks from scratch meant we inherited a principled difficulty progression and broad concept coverage. Our work sits on top of that: the Qiskit translation, the evaluation infrastructure, and the empirical analysis.

What we found

To show the benchmark does its job, we evaluated 16 models across 7 prompting configurations — 39,200 runs in total. A few things stood out.

It separates models cleanly. Best-configuration pass rates run from 32.3% to 83.1%, with frontier models averaging about 26 points above open-source ones. That’s a wide enough spread to be useful, with no ceiling or floor effects crowding the results together.

Implementing a known algorithm is very different from encoding a problem. Models do well when the task is to implement something documented — Simon’s algorithm sits at 82.1%, basic gates at 81.6%. They struggle when they have to cast a classical problem into quantum terms — encoding SAT for Grover’s search drops to 34.4%. The dominant failure mode across the board is logic errors, not syntax: the code runs and uses Qiskit correctly but produces the wrong quantum state.

Chain-of-thought is not a free win. It helped a few models — notably the reasoning-tuned ones — but degraded performance for the majority, leaving it mid-pack on average and behind simple few-shot prompting. The practical takeaway is that prompting strategy should track a model’s provenance rather than defaulting to “more reasoning is better.”

Get it

Everything is open. The dataset, the evaluation framework, and all the baseline results are released so anyone can reproduce the numbers or run their own models against the benchmark.

Paper:
Dataset:
Code:

If you work on quantum code generation, model evaluation, or quantum education tooling, I’d be glad to hear what you make of it.

We Got AI Agents to Train RL Models for Quantum Transpilation

Thu, 22 Jan 2026 00:00:00 +0000

What if you could tell an AI assistant: train a model of X qubits to synthesize linear function circuits using the topology YYYY and 100 million steps - and it just does it? That’s now possible with the new qiskit-gym-mcp-server.

Inspired by the work Hugging Face has done with - where Claude and other agents can submit fine-tuning jobs through natural language - we’ve built something similar for quantum circuit synthesis. The key difference: instead of fine-tuning LLMs, we’re training RL agents that learn to synthesize quantum circuits optimized for real hardware.

Your browser does not support the video tag.

What Is qiskit-gym-mcp-server?

The is a community MCP (Model Context Protocol) server that integrates the library. It exposes reinforcement learning-based quantum circuit synthesis as tools that AI agents can invoke autonomously.

The server enables three main synthesis capabilities:

Permutation synthesis: Uses SWAP gate routing to handle qubit permutations on constrained topologies
LinearFunction synthesis: Synthesizes CNOT circuits for linear boolean functions
Clifford synthesis: Generates Clifford circuits using H, S, and CNOT gates

How It Works

The workflow is straightforward. An AI assistant connected via MCP can:

Create environments for different quantum circuit problems, supporting various hardware topologies including linear arrays, grid layouts, and layouts from real IBM backends
Train models asynchronously using PPO or AlphaZero algorithms with real-time progress monitoring via TensorBoard
Save and manage trained models for later use
Synthesize circuits using the trained models

AI Assistant → MCP Client → create_*_env_tool → start_training_tool
 ↓ ↓
 gym_core.py (envs) training.py (RLSynthesis)
 ↓
 synthesize_*_tool
 ↓
 Optimized circuit (QPY format)

The Agentic Example

To demonstrate the full potential, we’ve included a LangChain-based agent example in the . The agent can:

Create RL environments for specific synthesis tasks
Configure and launch training runs
Monitor progress and adjust hyperparameters
Use trained models to synthesize optimized circuits
Save models for deployment

This isn’t a toy demo - it’s a proof of concept of how we envision training our AI-based transpilation models that can outperform most existing heuristic methods, as demonstrated in (Nature Computational Science, 2025).

Getting Started

Install with pip:

# Full installation with all servers
pip install qiskit-mcp-servers[all]

# Or just the gym server
pip install qiskit-mcp-servers[gym]

Configure your MCP client (Claude Code, Cursor, etc.) to connect to the server, and you’re ready to start training RL models through conversation.

Why This Matters

The intersection of AI agents and quantum computing opens up new possibilities for automating complex optimization tasks. Instead of manually configuring training runs and tuning hyperparameters, you can describe what you want in natural language and let the agent handle the details.

This is just the beginning. The same pattern can extend to other quantum computing tasks and other fields.

Links:

Pull Request:
qiskit-gym-mcp-server:
PyPI:
qiskit-gym:
HF Skills Training (inspiration):

Also shared on on January 23, 2026

Certified as Senior Software Engineering Manager at IBM

Fri, 09 Jan 2026 00:00:00 +0000

Celebrating my new certification! After a few months acting as an in-country Senior Manager, I got certified for that role by IBM.

Originally shared on

Run Qiskit Code Assistant Locally easily!

Tue, 02 Dec 2025 00:00:00 +0000

Don’t have access to the Qiskit Code Assistant because you’re not on an IBM Quantum Premium plan?

No worries—you can run it directly on your computer! We’ve recently improved the local setup experience. With our scripts, you can:

✅ Download the LLM optimized for local enviroments (GGUF)
✅ Enable inference on your machine (through Ollama)
✅ Automatically configure the Qiskit Code Assistant VSCode or JupyterLab extension

Give it a try and share your feedback—we’d love to hear how you use it!

More info:

#Qiskit #QuantumComputing #AI #Ollama #OpenSource #AIforQuantum

Originally shared on on December 2, 2025

Seeing AI Through the Curry–Howard Lens

Tue, 02 Dec 2025 00:00:00 +0000

The draws a deep equivalence between logic and computation: propositions as types, proofs as programs. It’s an elegant insight from the 1960s—and surprisingly relevant to where AI is headed.

Many researchers have explored formal verification, type-driven development, and reasoning-based generation in AI. But Curry–Howard offers something more: a conceptual framework for understanding what’s missing in current LLMs and what becomes possible if we push further.

The gap is narrowing. We’re already seeing glimpses of this future: reasoning models that decompose problems step-by-step, agents that iterate and verify their outputs, post-training techniques that reward logical consistency over mere plausibility. But these are still fundamentally pattern-matching systems with reasoning-like behavior grafted on. Curry–Howard suggests a more fundamental shift—one where programs are proofs and types are logical propositions about correctness.

What would the next leap look like?

Programs as proofs in scientific computing. Imagine an LLM generating not just a quantum circuit in Qiskit, but a circuit with accompanying correctness guarantees—proving it implements the intended algorithm, respects hardware constraints, or satisfies specific properties. In science and quantum computing, this means code we can trust, not just code that runs and passes tests.

Types as reasoning scaffolds. Dependently-typed languages like Coq or Lean already embody Curry–Howard: types express logical specifications, and programs that type-check are valid proofs. Current LLMs can generate code in these languages, but they don’t truly reason within them. The next generation should navigate proof spaces natively, generating solutions that are provably correct by construction, not by post-hoc verification.

AI as a verification partner. In quantum computing, bugs aren’t just inconvenient—they’re catastrophic. A misconfigured gate sequence, an incorrect assumption about decoherence, and months of experimental time are wasted. Today’s coding agents can catch some errors through testing and iteration. Tomorrow’s should provide formal guarantees.

The challenge ahead

The challenge is architectural: current approaches layer reasoning on top of statistical foundations. What we need are systems where formal reasoning is intrinsic—models that learn not just to predict tokens, but to construct and verify proofs, where the training objective itself rewards logical soundness.

This isn’t distant future. Research on neural theorem proving, program synthesis with specifications, and LLMs integrated with proof assistants is accelerating. What Curry–Howard offers is a north star: a vision of AI that doesn’t just mimic reasoning, but embodies it.

For fields like quantum computing, and science in general, where correctness is non-negotiable, this evolution from reasoning-flavored generation to proof-native systems could be transformative.

Note: The featured image for this post was generated using Gemini Nano Banana.

Qiskit Code Assistant Upgraded to Mistral-Small-3.2-24B-Qiskit

Thu, 27 Nov 2025 00:00:00 +0000

We’ve upgraded the Qiskit Code Assistant! 🚀

Last month, we introduced mistral-small-3.2-24b-qiskit, replacing granite-3.3-8b-qiskit.

Key Improvements

✅ Better accuracy: The new model ranks higher across key benchmarks, including QiskitHumanEval, QiskitHumanEval Hard, HumanEval, MathQA, ASDiv, SciQ and others (see the for more info).

✅ Improved UX: More precise responses for quantum programming tasks.

✅ Open access: Available on Hugging Face → .

This update is part of our mission to make quantum development more intuitive and efficient. 🌟

Explore it, test it, and share your feedback!

#Qiskit #QuantumComputing #AI #OpenSource #HuggingFace

Originally shared on on November 27, 2025

qiskit-ibm-transpiler Ranks #3 in Unitary Foundation 2025 Survey

Mon, 10 Nov 2025 00:00:00 +0000

🚀 🚀🚀 Climbing the ranks in quantum software!

I’m thrilled to share that the qiskit-ibm-transpiler library has moved up to #3 in the . Just last year, we — and this year’s leap is a testament to the incredible work of the team behind it.

In less than two years, the package has surpassed , becoming a relevant tool for quantum circuit optimization/transpilation across the community.

One of the most exciting updates this year: our AI transpiler passes — including the latest Pauli Networks pass ( ) — can now run in local mode, removing the need for an IBM Quantum premium plan. This makes advanced AI-powered transpilation more accessible than ever.

The Unitary Fund survey is widely recognized as the primary pulse check for quantum software adoption and developer preferences. Seeing our work reflected there is both humbling and energizing.

Huge kudos to the amazing team behind this David Kremer, Víctor Villar Pascual, Jesús Talavera Gómez, Yaiza García Martín-Mantero, Ismael Faro — your dedication, creativity, and deep technical insight are what make this possible. And thank you to the community for your continued trust and feedback.

Let’s keep pushing the boundaries of quantum + AI software together.

#QuantumComputing #Qiskit #IBMQuantum #AITranspiler #UnitaryFund #QuantumSoftware #OpenSource #AIforQuantum

Originally shared on on November 10, 2025 - 50 reactions as of 11/11/2025

David Peral García Successfully Defends His PhD Thesis on Quantum NLP

Mon, 29 Sep 2025 00:00:00 +0000

🎓 Last Friday was a very special day - David Peral García successfully defended his PhD thesis!

This milestone holds special significance for me as my first doctoral thesis as an advisor, and I couldn’t be prouder of David’s groundbreaking work.

His research on Quantum Natural Language Processing represents what is likely the first PhD thesis in Spain dedicated to this emerging field, positioning him among the pioneers shaping the future of quantum computing and NLP.

David’s contributions have already made a substantial impact in the research community, with notable publications including:

📄 A comprehensive survey on Quantum NLP in Computer Science Review:

📄 Novel applications in Expert Systems with Applications:

And there are more exciting contributions on the way that will further advance this fascinating intersection of quantum computing and language processing.

My deepest gratitude to the evaluation committee Mario Piattini, Anupama Ray, and Roberto Theron for their time, expertise, and valuable feedback, and to the co-advisor Francisco José García-Peñalvo for their essential collaboration throughout this journey.

Congratulations, Dr. Peral-García! 🎉

The future of Quantum NLP looks brighter with innovators like you leading the way.

#PhD #QuantumComputing #NLP #Research #QuantumNLP #AcademicMilestone

Originally shared on on September 29, 2025 - 121 reactions, 14 comments as of 11/12/2025

Keynote Speaker at Metafuturo 2025: Quantum Computing + AI Convergence

Wed, 24 Sep 2025 00:00:00 +0000

🚀 Thrilled to have been part of Metafuturo 2025 as a keynote speaker!

I had the incredible opportunity to share insights on the convergence of Quantum Computing + AI at this forward-thinking event organized by Bizkaiko Foru Aldundia / Diputación Foral de Bizkaia and ATRESMEDIA.

The intersection of quantum technologies and artificial intelligence represents one of the most exciting frontiers in tech today. During my talk, I shared what we are doing in this field at IBM Quantum. From LLMs and agentic AI applications for quantum computing to AI-optimized quantum circuits, we’re witnessing the birth of computational capabilities that seemed like science fiction just a few years ago.

What struck me most during the event was the genuine curiosity and engagement from attendees across diverse industries. The talks and panels ranged from practical implementations to philosophical implications—exactly the kind of multidisciplinary dialogue we need as we navigate this quantum journey.

Huge thanks to the organizers for creating such a vibrant platform for discussing the technologies that will shape our tomorrow. Events like Metafuturo 2025 are essential for building bridges between cutting-edge research and real-world applications.

The future is quantum-enhanced, AI-powered, and closer than we think! 🔬⚡

#QuantumComputing #ArtificialIntelligence #Innovation #Metafuturo2025 #TechTrends #QuantumAI #FutureOfTech #IBM

Originally shared on on September 24, 2025 - 107 reactions, 7 comments as of 11/12/2025

IEEE Quantum Week in Albuquerque: AI and Quantum Computing at IBM Quantum

Sun, 31 Aug 2025 00:00:00 +0000

🚀 Excited to be attending IEEE Quantum Week in Albuquerque, New Mexico! I’ll be sharing some of the work we’re doing at IBM Quantum, especially at the intersection of AI and Quantum Computing. If you’re around and passionate about Quantum, AI, or both—let’s connect! Here’s where you can find me:

🧠 Today, Sunday, August 31 – 4:00PM Paper Presentation: “Quantum Processing Unit (QPU) Processing Time Prediction with Machine Learning”. We’ll explore how ML can improve QPU time prediction for quantum jobs—enabling smarter quantum workload scheduling.

🛠️ Monday, September 1 – 10:00AM Tutorial: “TUT13 — AI Methods for Quantum Circuit Optimization”. Together with David Kremer, we’ll demonstrate how Reinforcement Learning can be used to transpile and synthesize quantum circuits, using qiskit-ibm-transpiler and other tools developed by our team.

💡 Tuesday, September 2 – 12:00PM Live Demo at IBM Quantum Booth #400. I’ll demo our new AI-powered quantum computing development environment—designed to bridge the gap between quantum theory and practical implementation in a visual way.

If you’re attending and want to talk about AI for Quantum, Quantum+AI, or our work at IBM Quantum—feel free to reach out or stop by!

#IEEEQuantumWeek #IBMQuantum #QuantumComputing #AIforQuantum #QuantumPlusAI

Originally shared on on August 31, 2025 - 42 reactions, 2 comments as of 11/12/2025

Quantum Verifiable Rewards for Post-Training Qiskit Code Assistant

Sat, 30 Aug 2025 00:00:00 +0000

📄 We just released a new paper “Quantum Verifiable Rewards for Post-Training Qiskit Code Assistant” - and it’s been getting great attention from folks in the quantum field since day one. 🎉

🔍 What we built: A novel approach to train AI models that can write better quantum code using Qiskit. What makes this interesting:

✅ Quantum verification at the core - Instead of just hoping the AI-generated code works, we actually verify it runs correctly on real quantum systems ✅ Smart training pipeline - We created synthetic quantum problem-test pairs and used both Direct Preference Optimization (DPO) and Group Relative Policy Optimization (GRPO) to align our models ✅ Real quantum feedback - Our models learn directly from quantum hardware and systems rewards, ensuring the code they generate actually works in practice

The results: Our best model significantly outperforms existing leading open-source baselines on the challenging Qiskit-HumanEval-hard benchmark.

This work presents a solid contribution to making quantum programming more accessible through AI assistance. The approach of using quantum hardware verification to train coding assistants opens up promising directions for future research.

Proud of our team’s thoughtful work on this project: Nicolas Dupuis, Adarsh Tiwari, Youssef MROUEH, David Kremer, Ismael Faro. The intersection of AI and quantum computing continues to offer compelling research opportunities.

Paper:

#QuantumComputing #AI #MachineLearning #Qiskit #Research #AIforQuantum

Originally shared on on August 30, 2025 - 76 reactions, 0 comments as of 11/12/2025

AI Transpiler Passes Enable Quantum Volume Breakthrough on IBM Hardware

Wed, 27 Aug 2025 00:00:00 +0000

🚀 Great to see this progress on the quantum hardware side! 🚀

I’m also really proud of our work on the AI transpiler passes that enabled this advancement in quantum volume. When I started working with David Kremer, Ivan Duran Martinez, Hanhee Paik and Ismael Faro on these AI-powered passes, we were motivated to push the limits of transpilation to deliver the best results to our clients.

It’s incredibly rewarding to see how our efforts are making real impact, like what Jay Gambetta has highlighted here.

Thanks to the entire team involved in the AI transpiler passes - Víctor Villar Pascual, Jesús Talavera Gómez, Yaiza García Martín-Mantero, Ayushi Dubal, Sanjay Vishwakarma, and many others - this kind of progress is only possible through all of our effort! 🙌🙌🙌

Context In response to: celebrating IBM’s r3 beta QPU (ibm_pittsburgh) achieving quantum volume milestones of 1024 and 2048, utilizing AI-optimized transpilation circuits developed by our team.

Originally shared on on August 27, 2025 - 31 reactions, 0 comments as of 11/12/2025

Qiskit Code Assistant: New Open-Source LLM Models Released

Tue, 08 Jul 2025 00:00:00 +0000

🚀 We’re thrilled to announce the latest open-source LLM releases from the Qiskit Code Assistant team! 🚀

These new models bring enhanced capabilities and broader compatibility to quantum computing development.

What’s New:

✨ Qiskit 2.0 Compatibility - Our newer models are fully compatible with Qiskit 2.0 code, ensuring seamless development with the latest quantum computing framework

🧠 Enhanced Text Understanding - Significant improvements in general text comprehension and code generation capabilities

🔬 Expanding Our Horizons - We continue to leverage the newest stable versions of IBM Granite 3.2 and 3.3 models, while for the first time exploring beyond IBM Granite with the powerful Qwen2.5-Coder series

Available Models: • Granite 3.3 8B Qiskit (+ GGUF) • Granite 3.2 8B Qiskit (+ GGUF) • Qwen2.5-Coder 14B Qiskit (+ GGUF)

All models are available on Hugging Face and ready for local deployment!

🔗 Browse All Models: 🔗 Get Started:

We’d love to hear how you’re making the most of these models! Share your experiences and quantum development journey with us. 🌟

Big thanks to the team involved in this release, specially to Nicolas Dupuis, Adarsh Tiwari and Ismael Faro

#QuantumComputing #OpenSource #AI #LLMs #Qiskit #IBM #TechInnovation #QuantumDevelopment #QiskitCodeAssistant

Originally shared on on July 8, 2025 - 304 reactions, 16 comments as of 11/12/2025

Returning to the University of Salamanca: Sharing My PhD Journey

Sat, 07 Jun 2025 00:00:00 +0000

Yesterday was a special day returning to my alma mater, the University of Salamanca to share my journey with current PhD students.

Speaking about my PhD experience and the career path that followed felt like coming full circle. It was inspiring to connect with brilliant doctoral candidates who are navigating their own research journeys.

My main message? Make the most of your PhD years - they’re not just about the research, but about developing the critical thinking, resilience, and problem-solving skills that will serve you throughout your career.

The energy and curiosity of the audience brought back memories of my own PhD days, which were truly an adventure. Getting to hang out with fellow PhD students and other speakers made the experience even more enriching.

A heartfelt thank you to Universidad de Salamanca, the Escuela de Doctorado and Alicia G. for the invitation and for creating such valuable spaces for knowledge sharing and mentorship.

Here’s to supporting the next generation of researchers! 🎓

#PhD #UniversityOfSalamanca #Academia #Research #Mentorship #DoctoralStudies

Originally shared on on June 7, 2025 - 115 reactions, 1 comment as of 11/12/2025

Among IBM Giants and Pioneers: Reflections from IBM Tech 2025 in Singapore

Fri, 28 Mar 2025 00:00:00 +0000

Among 👁️🐝Ⓜ️ Giants and Pioneers: Reflections from IBM Tech 2025 in Singapore 🇸🇬

I had the privilege of attending IBM Tech 2025 in Singapore this week - an exclusive, invitation-only gathering of IBM’s top technical talent from around the world. As one of the selected participants, I’m humbled to have been recognized among IBM’s most innovative and performing technical minds.

The atmosphere was electric as we dove deep into the future of AI, quantum computing, and emerging technologies. What struck me most was not just the cutting-edge content of the presentations from colleagues and leaders such as Ruchir Puri, Suja Viswesan, Ric Lewis, Borja Peropadre, and others, but the quality of conversations happening between sessions. Exchanging ideas with brilliant colleagues across different technical disciplines provided invaluable perspective on where technology is headed.

Particularly enlightening were the discussions with my IBM Quantum and AI Research colleagues and the cross-pollination of ideas between different technical domains. These connections reinforced my belief that the most transformative innovations happen at the intersection of diverse expertise.

Beyond the technical exchanges, Singapore provided a stunning backdrop for cultural experiences that complemented our professional activities. The city’s blend of tradition and futuristic vision perfectly mirrored our discussions about building tomorrow’s technologies while honoring established principles.

Returning home energized and inspired, I’m grateful for both the recognition and the opportunity to contribute to IBM’s technical community. The future belongs to those who collaborate across boundaries, and IBM Tech 2025 demonstrated that our collective expertise is our greatest strength.

#IBMTech2025 #Innovation #TechnicalExcellence #QuantumComputing #AI #FutureTech #IBMLife #AIforQuantum

Originally shared on on March 28, 2025 - 96 reactions, 14 comments as of 11/12/2025

Pauli Network Circuit Synthesis with Reinforcement Learning Paper Published

Wed, 19 Mar 2025 00:00:00 +0000

Really happy to see the paper on arxiv. The described AI-powered transpiler pass for Pauli Networks has been available in the Qiskit Transpiler Service since last November 2024, as presented in the Quantum Developer Conference 2024 Check out the paper and the related documentation on how to use it

Context: In response to about the Reinforcement Learning-based synthesis pass for Pauli Networks (Clifford + arbitrary angle Pauli rotation circuits). The research paper Pauli Network Circuit Synthesis with Reinforcement Learning was presented at the American Physical Society March Meeting in Anaheim.

Originally shared on on March 19, 2025 - 16 reactions, 2 comments as of 11/12/2025

New Qiskit HumanEval Release: Qiskit 1.4 Compatibility and Benchmark Improvements

Fri, 14 Mar 2025 00:00:00 +0000

New Qiskit HumanEval release! 🚀

We’ve just updated to be compatible with the latest Qiskit 1.4 release! But that’s not all - our update also includes significant improvements to the benchmark, making it more robust and rigorous in terms of code execution tests to provide even more accurate and comprehensive evaluations of your LLM-generated quantum code

View the full changelog:

Want to know more about Qiskit HumanEval? Check out our paper

#qiskit #AIforQuantum #qiskit_humaneval

Originally shared on on March 14, 2025 - 26 reactions, 2 comments as of 11/12/2025

Qiskit Code Assistant Now Compatible with OpenAI Completions API

Mon, 10 Mar 2025 00:00:00 +0000

Exciting update! The Qiskit Code Assistant service now exposes compatible endpoints with OpenAI’s Completions API. This integration enables seamless usage via existing libraries (OpenAI, LiteLLM, etc.), making it easy to infuse Qiskit knowledge into your LLM pipelines.

Learn more:

#qiskit #QiskitCodeAssistant #IBMQuantum #openai

Originally shared on on March 10, 2025 - 43 reactions, 0 comments as of 11/12/2025

Granite-8B-Qiskit-RC-0.10: Updated Checkpoint Using Current Training Approach

Mon, 03 Mar 2025 00:00:00 +0000

🙌 Thanks to the feedback received, we’ve released a revised version of the granite-8b-qiskit-rc-0.10! Check out the improved checkpoint at

#Qiskit #QuantumComputing #AI #AIforQuantum #QiskitCodeAssistant

Context In response to: .

Originally shared on on March 3, 2025 - 22 reactions, 0 comments as of 11/12/2025

Shipping Granite-8B-Qiskit-RC-0.10: The Final Model of Current Training Era

Tue, 18 Feb 2025 00:00:00 +0000

🚀 Last Friday, we shipped something special: granite-8b-qiskit-rc-0.10, our latest revision of the LLMs that empower the Qiskit Code Assistant.

Trained on a significantly expanded Qiskit synthetic dataset, this release marks the end of an era - it’s our final model using the current training approach.

What’s next? We’re pivoting to newer Granite base models and starting to integrate other cutting-edge techniques. Stay tuned!

Check it out:

#Qiskit #QuantumComputing #AI #AIforQuantum #QiskitCodeAssistant

Originally shared on on February 18, 2025 - 126 reactions, 8 comments as of 11/12/2025

qiskit-ibm-transpiler Ranks #4 in Unitary Fund 2024 Survey

Tue, 17 Dec 2024 00:00:00 +0000

🎉 Exciting news! The qiskit-ibm-transpiler has been recognized as the “4th most used quantum computing development tool globally” in the 2024 Unitary Fund survey under the “Full-stack development platforms, compilers, and simulators” category.

What makes this particularly remarkable? The project has only been publicly available for just over a year!

🔍 What is the qiskit-ibm-transpiler?

It’s a library that combines Qiskit’s powerful heuristic algorithms with our cutting-edge AI transpiler passes. Think of it as a sophisticated translator that optimizes your quantum code to run more efficiently on real quantum hardware.

🌍 What this means:

This recognition reflects both the growing adoption of quantum computing and the community’s trust in AI-powered tools. It’s an acknowledgment of not just our technology, but of every developer who has contributed to and embraced this project.

🔗 Resources:

GitHub:
Documentation:

Thank you to everyone who has been part of this journey. Here’s to pushing the boundaries of quantum computing together! 🚀

#QuantumComputing #OpenSource #Qiskit #IBM #SoftwareDevelopment #AITranspilerPasses #QiskitIBMTranspiler #AIforQuantum

Originally shared on on December 17, 2024 - 79 reactions, 1 comment as of 11/12/2025

Quantum Developer Conference 2024: Showcasing AI-Powered Quantum Tools

Fri, 15 Nov 2024 00:00:00 +0000

Excited to be participating in the Quantum Developer Conference 2024 at the IBM Thomas J. Watson Research Center! It’s been fantastic networking with fellow quantum computing professionals and showcasing the latest developments at the intersection of AI and quantum computing.

We’re featuring exciting innovations including the Qiskit Code Assistant and our cutting-edge AI-powered transpiler passes. These tools represent significant steps forward in making quantum computing more accessible and efficient.

Learn more:

#QuantumComputing #AI #IBMQuantum #AIforQuantum

Originally shared on on November 15, 2024 - 97 reactions, 1 comment as of 11/12/2025

Honored to Receive IBM Master Inventor Recognition

Thu, 24 Oct 2024 00:00:00 +0000

I’m honored to share that I’ve received the ! 🎉

This distinction is awarded to employees who have “mastered the patent process, mentored broadly, added value to IBM’s portfolio and demonstrated sustained innovation leadership and service.”

It’s incredibly rewarding to be recognized for contributions that blend innovation with mentorship and service to the broader community. This achievement reflects not just individual effort, but the collaborative spirit of working with brilliant colleagues across IBM.

Thank you to everyone who has been part of this journey!

Originally shared on on October 24, 2024 - 137 reactions, 40 comments as of 11/12/2025

Introducing Qiskit Code Assistant: New Blog Post Published

Tue, 15 Oct 2024 00:00:00 +0000

Last week we published a blog post summarizing what is the Qiskit Code Assistant and how you can start to use it.

Following the recent launch, users are already leveraging the tool’s features. We’re actively developing improved models with enhanced capabilities that we’re planning to release as open source!

Read the blog post:

#qiskit #qiskitcodeassistant #opensource #ibmquantum

Originally shared on on October 15, 2024 - 51 reactions, 0 comments as of 11/12/2025

IEEE Quantum Week 2024: Showcasing AI-Powered Quantum Tools

Tue, 24 Sep 2024 00:00:00 +0000

Las week I attended to the IEEE Quantum Week and it was a blast! 🚀🔬

I had the opportunity to witness and showcase several groundbreaking developments from IBM Quantum AI:

🔹 Qiskit Transpiler Service: Jay Gambetta presented impressive results showing how our AI-powered optimization passes outperformed all competitors in benchpress regarding depth and number of gates, including Qiskit itself!

🔹 Qiskit Code Assistant: We announced the first preview release of this innovative tool.

🔹 Unitary Compilation Research: David Kremer presented our work on approximate compiling of unitaries in quantum circuits using AI. We also co-delivered a tutorial on AI-powered transpiler passes.

🔹 Qiskit HumanEval: I had the chance to present this benchmark for evaluating quantum software development tools.

It was an incredible week of innovation, collaboration, and pushing the boundaries of what’s possible in quantum computing!

#QuantumComputing #IEEEQuantumWeek #Qiskit #AIforQuantum #QisktiCodeAssistant #QiskitTranspilerService #ResearchHighlights

Originally shared on on September 24, 2024 - 71 reactions, 0 comments as of 11/12/2025

Presenting Qiskit HumanEval at IEEE Quantum Week 2024

Wed, 18 Sep 2024 00:00:00 +0000

Hey! If you are at the #IEEEQuantumWeek 2024, I will be talking about the Qiskit HumanEval benchmark for LLMs in the session SYS-BNCH: Benchmarking (10:00 AM – 11:30 AM EDT)

After that, I’ll be around the IBM Quantum booth to chat about our AI and quantum computing initiatives. Looking forward to connecting with the community!

Originally shared on on September 18, 2024 - 51 reactions, 0 comments as of 11/12/2025

AI Methods for Approximate Compiling of Unitaries Paper Published

Wed, 07 Aug 2024 00:00:00 +0000

Excited to share our latest research: “AI methods for approximate compiling of unitaries”! 🚀

This work explores how artificial intelligence can make quantum circuit compilation more efficient. We focus on superconducting quantum hardware using fixed two-qubit gates and single-qubit rotations.

🔍 Our approach:

Three-stage process: identifying templates, predicting parameters, and refining through gradient descent
Uses deep learning and autoencoder-like models to suggest initial templates and parameter values
Demonstrates improvements over exhaustive search and random initialization on 2 and 3-qubit unitaries

This research highlights AI’s potential to enhance quantum circuit transpiling, supporting more efficient quantum computations on current and future hardware.

The paper has been accepted at QCE24 (Fifth IEEE International Conference on Quantum Computing and Engineering)!

Read the paper:

Originally shared on on August 7, 2024 - 58 reactions, 0 comments as of 11/12/2025

Qiskit HumanEval: Evaluation Benchmark for Quantum Code Generation Published

Wed, 03 Jul 2024 00:00:00 +0000

Excited to share our new research paper introducing the Qiskit HumanEval dataset! 🚀

This work addresses a critical need: evaluating Large Language Models’ capability to generate quantum computing code. Our dataset comprises more than 100 quantum computing tasks, each with accompanying prompts, solutions, test cases, and difficulty ratings.

We systematically tested LLMs on their ability to produce executable quantum code, demonstrating the feasibility of using generative AI tools in quantum code development and establishing important benchmarks for the field.

This research opens new possibilities for AI-assisted quantum software development and provides a standardized way to measure progress in this exciting intersection of quantum computing and artificial intelligence.

Read the paper:

#quantumcomputing #ibmquantum #qiskit #llms

Originally shared on on July 3, 2024 - 70 reactions, 7 comments as of 11/12/2025

Starting New Role: First Ever AI for Quantum Product Owner at IBM Quantum

Thu, 27 Jun 2024 00:00:00 +0000

I’m happy to share that I’m starting a new position at IBM Quantum as the 1st ever AI for Quantum Product Owner! So happy and motivated for what is coming 🚀

Originally shared on on June 27, 2024 - 182 reactions, 52 comments as of 11/12/2025

IBM Develops The AI-Quantum Link: Featured in Forbes

Tue, 25 Jun 2024 00:00:00 +0000

Excited to share this ! 🚀

The integration of AI and Quantum Computing has the potential to transform industries and advance quantum computing capabilities significantly. We’re thrilled to be disclosing our results and research papers from this exciting field.

This represents a major step forward in making quantum computing more accessible and powerful through the application of artificial intelligence.

Context: Sharing about IBM Quantum’s work integrating AI with quantum computing.

Originally shared on on June 25, 2024 - 98 reactions, 3 comments as of 11/12/2025

Optimize Quantum Circuits with AI-Powered Transpiler Passes

Mon, 17 Jun 2024 14:00:00 +0000

Exciting updates from our team at the convergence of AI and quantum computing! 🚀

We have released a bunch of new features in the Qiskit transpiler service and the Qiskit Code Assistant projects that continue unlocking the potential of applying classical AI into Quantum computing.

Recent releases include: 🔹 Beta version of the Qiskit transpiler service (unveiled at #THINK24) 🔹 Research paper on AI methods enabling AI-powered transpiler passes 🔹 Qiskit Code Assistant with accompanying paper introducing LLMs and the Qiskit HumanEval benchmark

Both projects have received recognition through the IBM Quantum Challenge, with fantastic feedback from participants who got to experience the power of AI-driven quantum optimization firsthand!

This work represents a major step forward in making quantum computing more efficient and accessible through the integration of artificial intelligence.

Original post:

Originally shared on on June 17, 2024 - 39 reactions, 1 comment as of 11/12/2025

Qiskit Code Assistant: Training LLMs for Quantum Code Generation Paper Published

Mon, 17 Jun 2024 12:00:00 +0000

Excited to share our research paper on training specialized LLMs for quantum computing code generation using Qiskit! 🚀

Code Large Language Models have emerged as powerful tools, revolutionizing the software development landscape by automating coding tasks. However, quantum programming presents unique challenges compared to classical coding.

Our work addresses:

The scarcity of quantum code examples
The rapidly evolving nature of the quantum computing field
Custom benchmarking similar to HumanEval for quantum-specific tasks

The results are promising: our model outperforms existing state-of-the-art quantum computing models!

This research represents an important step toward making quantum programming more accessible through AI assistance.

Read the paper:

#qiskit #genAI #codegen #LLMs #IBMQuantum

Originally shared on on June 17, 2024 - 45 reactions, 3 comments as of 11/12/2025

Reinforcement Learning for Quantum Transpiling: Research Paper Published

Mon, 17 Jun 2024 10:00:00 +0000

Excited to share research demonstrating the integration of Reinforcement Learning (RL) into quantum transpiling workflows for the Qiskit transpiler service! 🚀

This work achieves near-optimal circuit synthesis and routing with significant performance improvements over traditional optimization methods like SAT solvers.

Key achievements: ✅ Linear Function, Clifford, and Permutation circuit synthesis up to 65 qubits ✅ Substantial reductions in two-qubit gate depth for routing up to 133 qubits ✅ Performance advantages over SABRE routing heuristics ✅ Practical efficiency for quantum transpiling pipelines

This research represents a major step forward in making quantum computing more efficient and accessible through AI-powered optimization.

Big thanks to the amazing team: David Kremer, Víctor Villar Pascual, Hanhee Paik, Ivan Duran Martinez, and Ismael Faro!

Read the paper:

#qiskit #quantumcomputing #IBMQuantum

Originally shared on on June 17, 2024 - 48 reactions, 0 comments as of 11/12/2025

Comparing Natural Language Processing and Quantum NLP: Research Published

Thu, 06 Jun 2024 08:00:00 +0000

Excited to share our peer-reviewed publication “Comparing Natural Language Processing and Quantum Natural Processing approaches in text classification tasks”! 🚀

This collaborative work with David Peral García and Francisco José García-Peñalvo has been published in Expert Systems with Applications.

Key findings:

✅ Quantum NLP models can obtain the same or better results in some datasets for simpler text classification tasks like entity recognition and sentiment analysis
✅ Performance diminishes with increased label complexity and sentence difficulty
✅ Experiments utilized up to 7 qubits and tested on both new and existing datasets across multiple classification scenarios

While the results are promising, we identified that advancing to larger-scale quantum applications requires more powerful quantum hardware and sophisticated software infrastructure.

This research contributes to our understanding of where quantum computing can provide advantages in natural language processing tasks and highlights the path forward for future developments.

Read the full paper on .

Originally shared on on June 17, 2024 - 19 reactions, 0 comments as of 11/12/2025

Qiskit SDK v1.0 Released: Including Qiskit Transpiler and Code Assistant

Thu, 16 May 2024 00:00:00 +0000

I’m extremely proud of this massive release including two projects from my team: the Qiskit Transpiler (with the AI transpiling passes) and the Qiskit Code Assistant 🚀

Qiskit SDK 1.x represents full-stack software for all things quantum, featuring:

Qiskit Transpiler with AI-powered optimization
Qiskit Runtime
Qiskit Serverless
AI Code Assistant

More updates and info coming soon!

Context: In response to , introducing the full-stack software for quantum computing with Qiskit Transpiler, Qiskit Runtime, Qiskit Serverless, and the AI Code Assistant.

Originally shared on on May 16, 2024 - 15 reactions, 0 comments as of 11/12/2025

Systematic Literature Review: Quantum Machine Learning and Its Applications Published

Mon, 05 Feb 2024 00:00:00 +0000

Excited to share our systematic literature review paper “Systematic literature review: Quantum machine learning and its applications” published in Computer Science Review! 🚀

This comprehensive work, in collaboration with David Peral García and Francisco José García-Peñalvo from the University of Salamanca, Spain, analyzes the state of quantum machine learning research from 2017 to 2023.

Key findings from our review of 94 studies:

✅ Identified two primary algorithm categories: quantum versions of classical ML algorithms (support vector machines, k-nearest neighbors) and quantum neural networks
✅ Image classification emerged as a particularly relevant application area
✅ While quantum machine learning demonstrates promise, it remains far from achieving its full potential

Our analysis highlights that quantum hardware improvements are necessary, as current quantum computers lack sufficient quality, speed, and scalability for QML’s full realization.

This research provides valuable insights into the current state and future directions of quantum machine learning.

Read the full paper: .

Originally shared on on February 5, 2024 - 53 reactions, 4 comments as of 11/12/2025

IBM Quantum Recruiting AI Engineer Interns in Spain

Sat, 03 Feb 2024 00:00:00 +0000

We’re recruiting AI Engineer interns based in Spain for the IBM Quantum team! 🚀

We are introducing new AI-based capabilities in our software stack, including:

AI circuit transpilers compatible with Qiskit
LLM-powered code assistants

As an intern, you would support:

AI model training and deployment
Software service development
MLOps work

We’ve already received over 100 applications, and our team of international engineers is welcoming to newcomers. This is a fantastic opportunity to gain practical AI experience combined with quantum computing knowledge!

Apply here:

#AIEngineer #internship #Spain #IBMQuantum

Originally shared on on February 3, 2024 - 15 reactions, 0 comments as of 11/12/2025

IBM Quantum Summit 2023: Major Announcements on Quantum Utility

Thu, 07 Dec 2023 00:00:00 +0000

I am thrilled and honored by the major announcements from IBM Quantum Summit 2023! 🚀

Key highlights advancing quantum utility:

Hardware Breakthroughs:

IBM Heron chip: 3-5x device performance improvement over Eagle
IBM Condor: 1,121-processor system demonstrating scaling solutions
IBM Quantum System Two: Now operational at Yorktown Lab with 3 Heron processors

Software Innovations:

Qiskit 1.0: Scheduled February release with improvements in circuit construction, compilation times, and memory consumption
Quantum Serverless: Beta deployment for scaled pattern execution
AI Integration: Automated code development via watsonx and enhanced transpiler tools

Vision: Extended roadmap through 2033

Academic and industry partners from University of Tokyo, Argonne National Lab, BasQ, and others showcased utility-scale quantum applications.

This represents tremendous progress toward realizing the full potential of quantum computing!

Context: In response to , highlighting hardware and software advances toward quantum utility.

Originally shared on on December 7, 2023 - 6 reactions, 0 comments as of 11/12/2025

Starting New Role: Software Engineering Manager

Sat, 06 May 2023 00:00:00 +0000

I’m happy to share that I’m starting a new position as Software Engineering Manager!

Originally shared on on May 6, 2023 - 85 reactions, 10 comments as of 11/12/2025