ResolveAI today introduced SRE-1, a model trained to find the root cause of production alerts. On hard production incidents, SRE-1 dominates the Pareto frontier of pass rate against cost. Measured sep...

SRE-1 is the most accurate model at every budget ResolveAI measured on hard production incidents. Its agents now go on the critical path, starting with Primary On-call.
SAN FRANCISCO: ResolveAI today introduced SRE-1, a model trained to find the root cause of production alerts. On hard production incidents, SRE-1 dominates the Pareto frontier of pass rate against cost. Measured separately as a complete system, ResolveAI scores higher on production investigations than the coding-agent harnesses tested against it, with the widest margin on the hardest alerts. ResolveAI brings this research into its SRE agents that can now go on the critical path, starting with on-call, to drive reliability improvements. ResolveAI Primary On-call takes the page, closes what it can, and escalates to engineers only when needed, with an investigation attached. It also goes after the underlying issues that keep generating those pages.
AI has transformed how software gets written and barely touched how it gets run. Models learned to write code from billions of published lines, whereas production offers no comparable public training corpus: no two companies run the same system, incidents stay private, root causes are never written down, and the evidence an investigation depends on expires within weeks. General-purpose models, however capable, arrive at an incident without knowledge of the environment they are asked to investigate, and a larger model does not change that, because the answer was never written down anywhere. It exists only in the running system, reachable by making the right queries against that system, in the right order, and judging what comes back before the evidence ages out. That is work a domain-specific model for production has to be trained for, not knowledge it can be handed from the internet.
"No engineering team can keep up with how fast code and patches land in production. The only answer is agents accurate, fast, and cheap enough to put on the critical path, and that takes a model built for production. On hard production incidents, nothing we tested is more accurate at the same price than SRE-1, frontier models included, so now every alert gets investigated and the problems behind them get fixed," said Spiros Xanthos, co-founder and CEO of ResolveAI.
SRE-1 is trained for production
SRE-1 is a 27B model built on open weights, trained by a team of 16 AI researchers over a final run of 7.4 days on up to 512 GPUs. On 49 difficult live production cases, each run five times, an ensemble of five SRE-1 investigations matches the strongest frontier model the company tested, at lower cost. Training used reinforcement learning alone, which took a single investigation's pass rate from 6.9% to 38.8% on the same incidents with the same tools. Every tool in training was pinned to the moment the alert fired, so the model could not read the answer from the future.
"ResolveAI's SRE-1 model shows what a domain-specific model can deliver for production. The fact that ResolveAI has managed to produce this model is a testament to the team's deep modeling and RL talent. We are proud to have supported ResolveAI as the first design partner for Snowflake Cortex Training, which they used to build SRE-1," said Vivek Raghunathan, SVP of Engineering at Snowflake.
ResolveAI benchmarks the complete system separately. On 178 production alert investigations graded against verified root causes, Resolve solved about twice as many incidents as the general-purpose coding agents run on the same alerts. On the hardest third of those alerts, Resolve's root-cause score was 15 to 25% higher.
"AI in production has to deliver more than promising results in a demo. It needs to be accurate, dependable, and economically viable at scale. Domain-specific models like ResolveAI's SRE-1 represent an important step toward making AI a meaningful part of how we detect, investigate, and resolve incidents. At Zscaler, where we protect the world's largest enterprises, including more than 50% of the Fortune 500, we see significant opportunity to work with ResolveAI to improve reliability for our customers," said Misha Kuperman, Chief Reliability Officer & GM, Zscaler.
ResolveAI is on the critical path for every page
Primary On-call holds the pager: it takes the page, closes what it can, and escalates only when an engineer is needed. On each page, it investigates across telemetry, infrastructure, code, and operational knowledge. It then closes the alert, mitigates the problem where a team's policies permit, or escalates when human judgment is needed. After it acts, it checks the fix against the production baseline and keeps watching until recovery holds. In production, Primary On-call handles up to 80% of paged alerts without escalating to a human, and customers report a 10x improvement in mean time to resolution, with up to 95% root-cause accuracy on investigations.
Existing paging-system timeouts remain an independent backstop, and Resolve records every investigation and decision so engineers can review what it found and why it acted.
"ResolveAI has been part of how our CRM Analytics team handles incidents, so extending it to ResolveAI Primary On-call was the natural next step. In the first few weeks of our dark launch testing, ResolveAI has closed more than 70% of our paging alerts without waking an engineer, with high accuracy. That's why we're ready to put ResolveAI Primary On-call in our paging path as our first responder," said Bhupesh Patel, Software Engineering Principal Architect at Salesforce.
Addressing underlying issues to prevent alerts and incidents
An alert that is answered but never fixed keeps firing, and so does every other alert the same underlying problem produces. Fixing that problem stops the alerts and removes a source of future incidents.
Primary On-call connects related alerts and investigations into a shared view of each underlying problem. Across a team's services, those views add up to a running picture of service health that the team can keep improving. It carries the production evidence and investigation history into the fix, applies it within the scope a team has approved, then monitors it to confirm it holds. The evidence and the full audit trail stay with the issue.
Expanding autonomy as trust grows
On the alerts a team knows best, Resolve works every page and records the call it would have made while the existing rotation continues unchanged, so engineers can grade its judgment against pages they already know the answer to. Primary On-call takes the rotation only on the alert rules where that record has earned it.
Primary On-call is generally available today. SRE-1 is available only through ResolveAI.
For more information, visit resolve.ai.
About ResolveAI
ResolveAI is the AI Lab building models and SRE agents that automate production. Founded by Spiros Xanthos and Mayank Agarwal, co-creators of OpenTelemetry, it trains its own models, including SRE-1, and puts its agents on the critical path: taking the pager, mitigating incidents, fixing underlying issues, and automating the day-to-day work of production. Engineering teams at Coinbase, DoorDash, Figma, MSCI, Nasdaq, Salesforce, SAP, and Zscaler run production with ResolveAI. Learn more at resolve.ai.
Fonte: Business Wire
Alaa Abdul Nabi, Vice President, Sales International at RSA presents the innovations the vendor brings to Cybertech as part of a passwordless vision for…
G11 Media's SecurityOpenLab magazine rewards excellence in cybersecurity: the best vendors based on user votes
Always keeping an European perspective, Austria has developed a thriving AI ecosystem that now can attract talents and companies from other countries
Successfully completing a Proof of Concept implementation in Athens, the two Italian companies prove that QKD can be easily implemented also in pre-existing…
Ahead of Anthropic’s upcoming initial public offering, the Center for AI Safety (CAIS) today joined a rare coalition of prominent figures in signing the…