Corti S1 just beat OpenAI’s GPT-6.2 Sol and Astra on speed and cost in six real-world coding tasks. 🏁
A few days ago, our Senior Platform Engineer, Chituru Chinwah, put an open-weight AI model running on European sovereign infrastructure to the test against OpenAI’s GPT-6.1 Sol. Both models fixed all six bugs, but Corti S1 finished 1.8× faster.
So, when OpenAI released its latest models, of course we had to run the same experiment again to see how they stacked up. Same bugs, same tools, same rules.
Corti S1 won five of six rounds against GPT-6.2 Sol, finishing 1.2× faster. Against GPT-6.2 Astra, S1 won four of six rounds, finishing 1.1× faster while costing 73% less.
Why does this matter? It demonstrates the potential of open-weight models running on European sovereign infrastructure to deliver competitive performance at a lower cost, while giving businesses greater control over their AI.
Watch how they unfolded: corti.ai/blog/corti-s1-gpt-6…
1
74
Speech recognition has come a long way, but getting clinical speech right is still far from a solved problem.
Medical terminology, drug names, dosages, contextual ambiguity and clinical shorthand continue to challenge even the most advanced AI models.
Our research published in May demonstrated why speech recognition needs to be evaluated against the specific demands of clinical workflows.
As healthcare moves towards autonomous AI agents, getting these foundations right becomes even more important. Accurately capturing what clinicians say is just the beginning. Understanding what they mean and reliably translating that intent into action is the next frontier.
Daniel Engel explores the evolution from dictation to ambient AI to conversational agents, and why we can't afford to skip steps along the way.
Read the full article on @hitconsultant hitconsultant.net/2026/11/06…
1
1
46
What does it take to build AI for healthcare?
Many of the most important engineering decisions happen long before a model reaches production. From the infrastructure it runs on to where data is processed, these choices shape how AI can be deployed and who has control over it.
Our Director of Platform Engineering, Nicklas Frahm, caught up with @TechAllyson from TechArena.ai to discuss how we're building these foundations at Corti.
From building our own Kubernetes engine to sovereign AI deployment and embracing open-source technology, Nicklas shares the engineering thinking behind our approach and what it takes to give organisations greater control over their AI.
Read the full interview: techarena.ai/content/corti-o…
1
67
Big news for healthcare AI in France!
Corti is bringing its sovereign healthcare AI offering to France, powered by Scaleway’s European cloud infrastructure.
For developers building healthcare solutions, this means access to high-performing, clinical-grade AI with the ability to keep sensitive patient data and AI workloads within Europe, under European jurisdiction.
And there’s more to it than where data lives, developers get greater choice over the models they use, where they run and how they scale.
An exciting step towards making advanced, sovereign healthcare AI more accessible, without compromising on the performance and flexibility required for clinical AI.
Read more here lnkd.in/gsg9hBZy
42
Europe has the talent, expertise and ambition to build world-class AI. The next step is taking greater control of the technology that powers it.
That’s what our CEO and Co-founder, @andreascleve , discussed in his latest interview with @borsendk.
“Large Danish and European companies need to start owning their own AI destiny.”
That ambition is driving our next chapter. Building on a decade of experience in healthcare, we’re bringing our AI models, infrastructure and tools to more regulated industries, enabling organisations to build differentiated AI on their own terms.
Read the full interview here. borsen.dk/nyheder/tech/dansk…
48
Last chance to register! ⏳
For developers building and deploying AI in highly regulated environments like healthcare, the questions are getting more challenging.
How do we build AI agents that can be trusted in healthcare? What does digital sovereignty actually require? And how do we make AI systems more transparent?
This Thursday, we’re bringing these conversations to Corti HQ in Copenhagen.
We’ve also added two more speakers! Daniel Varab, Senior Machine Learning Engineer and Jakob Havtorn Speech Recognition Team Lead will join the session on High-Stakes AI in Europe.
Here’s the full agenda:
⚖️ EU AI Act in Healthcare: Sonia Moula
🔐 EU Digital Sovereignty: Nicklas Frahm
🔍 AI Transparency & Watermarking: Sonia Moula, LL.M. & Nicklas Frahm
🤖 AI Agents in Healthcare: Henrik Cullen
🧠 High-Stakes AI in Europe: Kevin Pelgrims, Jakob Havtorn & Daniel Varab
We’ll have live polls, audience challenges and open discussion. Come ready to share your perspective, challenge ideas and meet others working on the future of AI in Europe.
And yes, dinner, drinks and pizza are on us. 🍕
📍 Corti HQ, Copenhagen
🗓️ Thursday, 12 November | 17:00–21:00 CET
See you there!
Register 👉 luma.com/corti-eu-ai
Building an AI agent is one thing. Getting it into production raises a whole new set of questions.
We've expanded the Corti Agentic Framework with six capabilities to help developers address them, from connecting tools and tracing execution to evaluating performance.
👩💻 Explore the technical guide corti.ai/guides/what-is-new-…
42
In healthcare, we've learned that constraints can become an advantage. Building powerful AI models is one thing, but making them accurate, compliant and reliable requires depth. Our CEO @andreascleve Cleve on why we went deep, not broad, on the EUVC podcast.
🎧 eu.vc/podcast/andreas-cleve-…
46
We've been a little quiet 👀 but October took us across Europe.
7 cities. @larsmaaloe on sovereign AI and open-weight models with @Scaleway and @NVIDIA; @andreascleve on why agentic AI needs control @NVIDIAGTC.
Our takeaway? Models matter. What we build matters more. 🌍
And we've got plenty more to share. Stay tuned. 🚀
corti.ai/blog/october-wrappe…
4
351
Corti won 1st place in Digital Health at the 2026 Nebius AI Discovery Awards, selected from 647 companies worldwide. 🏆
Proud of the team building AI for one of healthcare’s hardest problems.
Thank you, @nebiusai.
Now, back to building.
We’re hiring: lnkd.in/eE5fZiWe
Hackathon Aug 19–21 @ Corti HQ
Pick a clinical problem. Build on our full API stack (speech-to-text, agents, medical coding). Ship by Friday.
$4,500 in prizes + Investor intros for 1st place.
Devs + clinicians welcome.
→ luma.com/corti-CPH-2026
→ corti.ai/events
3
109
⚫🔴🟡 Symphony for Medical Coding now supports German ICD-10-GM
One API call. ICD-10-GM inpatient + outpatient. Full audit trail.
Partner S3 Praxiscomputer is already live.
Start building: console.corti.app/auth
The best AI labs are racing to understand how models actually work. Today we're introducing GIM (Gradient Interaction Modifications), top-ranked on the @huggingface Mechanistic Interpretability Benchmark and open-sourced. 👏🏻Joakim Edin. Python package: bit.ly/4s21YkL
Two years into AI scribes. New clinical trials show real progress, but not yet true productivity gains.
Corti focuses on the vertical healthcare AI infrastructure that developers and EHRs use to power safer, more effective tools.
Full perspective: bit.ly/48NjMbz
157
🇩🇰 New: Corti’s clinical-grade Danish ASR is live.
Our new model delivers ~2× WER reduction and a large Medical Term Recall gain on clinical audio.
Read more → bit.ly/3JXvnv0
Live → docs.corti.ai
1
137
🚀 We’ve launched the largest French medical speech recognition model for healthcare 🇫🇷🏥🤖
Benchmarks show:
✅ +36% medical term recall
✅ 2.11% WER
✅ <1% CER
And in our latest demo: 99% medical term recall, 0 hallucinations. Allez!
2
151
Dan Engel tested Corti’s dictation accuracy on a long radiology report: 6% WER, 82% key-term match across 172 terms — strong medical speech recognition, clean unit handling.
Try it → docs.corti.ai/get_started/ge…
1
144
Healthcare apps are going voice-first - but most ASR is either flexible & unsafe or medical-grade & locked up.
At Corti, we’re rebuilding the ASR layer for healthcare: accurate, safe, open infra developers can actually build on.
bit.ly/433cQnC
1
101

