Summit recap

Write Quorlyn's recap email covering everything announced at the first Quorlyn Summit. Recipients are people subscribed to Quorlyn product updates.

The Summit ran across two days last week in Seattle. Quorlyn launched new models, hosted inference infrastructure, a self-serve agent platform, and a voice app. The recap landing page is https://quorlyn.s3.amazonaws.com/summit

Model updates:

• QS-4.1. Updated QS-4.1 variants for conversational and work-agent workloads. On Quorlyn's production benchmark of voice conversations, it leads Halcyon 5.4 Mini and Larkspur 3.5 Flash across voice-call capability, instruction following, tool calling, and linguistic quality, at $0.80 per million blended tokens. https://quorlyn.s3.amazonaws.com/qs-4-1
• Quorlyn Sight 2.0. Sight 1.0 launched in February and has since digitised more than 35 million pages. Sight 2.0 adds native key-value extraction from tables and forms, handwriting recognition, stronger complex-table parsing, and improved general OCR. https://quorlyn.s3.amazonaws.com/sight
• Quorlyn Transcribe 4. Speech recognition with five output modes from one model: verbatim transcription, normalised transcription, code-mixed output, transliteration, and translation. https://quorlyn.s3.amazonaws.com/transcribe
• Quorlyn Voice 4 (Research Preview). Quorlyn's most expressive text-to-speech model yet. It gives developers direct control over emotion, emphasis, pacing, and expression as speech is generated. https://quorlyn.s3.amazonaws.com/voice

Quorlyn Inference is the serving layer for Quorlyn models and frontier open-weight models, hosted from infrastructure in the United States and the EU. It gives teams data residency, governance, and the ability to route models on quality, latency, and cost. https://quorlyn.s3.amazonaws.com/inference

Quorlyn Atlas is one platform for all AI work: a single workspace, multiple agents. https://quorlyn.s3.amazonaws.com/atlas

• Voice Agents are generally available with no waitlist, after 325 million minutes of real conversations. Describe the goal in plain language, simulate a thousand calls, and be live on a phone number in under two hours. $0.06 per minute, everything included. https://quorlyn.s3.amazonaws.com/atlas/voice-agents
• Work Agents give an enterprise a fleet of AI coworkers that draft, research, analyse, and run processes alongside existing teams, under enterprise-grade security and compliance. Live now. https://quorlyn.s3.amazonaws.com/atlas/work-agents
• Content Agents bring AI voices, cloning, transcription, dubbing, and live translation into one studio, with dubbing that preserves emotion and produces lip-sync-ready audio. https://quorlyn.s3.amazonaws.com/atlas/content-agents
• Quorlyn Code is a coding agent for long-horizon tasks, started from an open-source coding agent and rebuilt for open models. The harness is worth 20.7 points on Data Agent Bench over the stock harness. Early beta. https://quorlyn.s3.amazonaws.com/code

Quorlyn Murmur is a voice app that lets people speak to their computer and get things done across any application. It understands more than 40 languages, including code-switching, and turns speech into clear, accurate text or action. Murmur can be tried at https://quorlyn.s3.amazonaws.com/murmur

Much of this is open to try now, and the rest rolls out in the weeks ahead. The full recap with benchmarks and demos is at https://quorlyn.s3.amazonaws.com/summit/recap and reading it is the email's one button; every other destination, including the recap landing page, is a text link.

Reference artwork: assets.illustrations.summitHero for the banner, summitIconQs41, summitIconSight, summitIconTranscribe, and summitIconVoice for the model updates, summitIconVoiceAgents, summitIconWorkAgents, summitIconContentAgents, and summitIconCode for Atlas, and summitMurmur for Murmur.

Use /world/quorlyn-email for brand chrome and CSS. Use the absolute HTTPS URLs from tokens.json; do not use /world/ paths as image src.

Dark mode should be supported.

Leave the HTML in output/email.html, a plain-text version in output/email.txt, and the subject in output/meta.json as {"subject": "..."}. Don't include From, To, or Date headers.