Resolve AI launches SRE-1, a 27B-parameter model trained to find root causes of production alerts
Resolve AI has launched SRE-1, a 27 billion-parameter model built on open weights and trained with reinforcement learning to find root causes of production alerts. The company also made Primary On-call generally available, putting its AI agents on call ahead of human engineers.
Resolve AI Inc. has launched SRE-1, a 27 billion-parameter model trained to find the root cause of production alerts, alongside Primary On-call, a service that puts its AI agents on call ahead of human engineers. Primary On-call is now generally available. Access to SRE-1 is limited to Resolve AI's own platform, SiliconANGLE reported.
Model details
- Size: 27 billion parameters, built on open weights.
- Training team: 16 AI researchers.
- Final training run: 7.4 days on up to 512 GPUs.
- Method: reinforcement learning alone.
- Training platform: Snowflake Inc.'s Cortex Training, a serverless service for customizing open-weight models. Snowflake Senior Vice President of Engineering Vivek Raghunathan said Resolve AI was the first design partner for the service.
Context window and pricing are not yet disclosed.
Resolve AI argues that AI has transformed how software gets written but has barely touched how it gets run. According to the company, production has no public training corpus comparable to the published code that models learned from, because incidents at each company stay private.
Benchmark results
All figures below come from Resolve AI.
- Measured on the same incidents before and after training, the pass rate for a single investigation rose from 6.9% to 38.8%.
- On 49 difficult live production cases, each run five times, an ensemble of five SRE-1 investigations matched the strongest frontier model tested at a lower cost. The company did not name that model in the announcement as reported.
- On 178 production alert investigations, Resolve AI's complete system solved about twice as many incidents as general-purpose coding agents.
These are company-run evaluations, not independent results.
Primary On-call
Primary On-call places the agent on the critical path for incoming pages. According to Resolve AI, each page is investigated across telemetry, infrastructure, code and operational knowledge. Depending on the findings, the agent can close the alert, mitigate the problem where a team's policies permit, or escalate to a human. After acting, it checks the fix against the production baseline and monitors until recovery holds. It also links related alerts and investigations into a shared view of each underlying problem.
For teams not ready to hand over the pager, a shadow mode records the call the agent would have made on each page.
Resolve AI says up to 80% of paged alerts in production are handled by Primary On-call without escalation. It also says customers have reported a tenfold improvement in mean time to resolution and root-cause accuracy of up to 95%.
Salesforce Inc. is among the early users. Bhupesh Patel, a software engineering principal architect at Salesforce, said that in the first few weeks of dark launch testing, the agent closed more than 70% of the CRM Analytics team's paging alerts "without waking an engineer."
Company background
Resolve AI is venture capital-backed and has raised more than $190 million to date. That includes a $40 million extension to its Series A in April, led by DST Global and Salesforce Ventures, which valued the company at $1.5 billion. Customers include Coinbase Global Inc., DoorDash Inc., Figma Inc., MSCI Inc., Nasdaq Inc., SAP SE and Zscaler Inc.
Chief Executive Spiros Xanthos said: "No engineering team can keep up with how fast code and patches land in production. The answer is agents that are accurate and fast enough to sit on the critical path, and that takes a model built for production."
What this means
Resolve AI is pairing a purpose-trained model with a product that acts on pages, rather than relying only on a general-purpose model. The reported training setup is small: 16 researchers, a 7.4-day final run and reinforcement learning on an open-weight base. The company's claim is that this was enough to move single-investigation pass rate from 6.9% to 38.8% on its own incident set. The ensemble result on 49 cases is the company's evidence for cost efficiency against a frontier model, though the model is not named and the evaluation is not independent. SRE-1 is available only inside Resolve AI's platform, so teams cannot call it directly. For operations teams, the shadow mode is the practical way to compare the agent's calls against human decisions before granting it authority to close or mitigate alerts.
Related Articles
RadixArk releases DaisyRose-27B, lifting Terminal-Bench 4.1 pass@1 from 5.6% to 12.0%
RadixArk released DaisyRose-27B, its first model, built on Qwen3.8-27B and trained with reinforcement learning on 1,200 curated agent tasks. According to RadixArk, it raises Terminal Bench 4.1 pass@1 from 5.6% to 12.0% and scores higher than the base model on all seven reported benchmarks. Weights ship under Apache 2.0 in BF16, FP8 and NVFP4.
SID AI releases SID-2 agentic search model; 12-agent Swarm scores 0.720 NDCG
SID AI has released SID-2, an agentic search model trained with reinforcement learning. A single agent reaches 0.585 NDCG in 3.11 seconds, and a 12-agent SID-2 Swarm scores 0.720 NDCG, which SID says is above GPT-6 Astra on its benchmark.
Tencent releases Hy4 flagship weights: 770B-parameter MoE, 1,048,576-token context, conditional license
Tencent released downloadable weights for its flagship Hy4 on November 20, with a System Card. The multimodal MoE has a 770-billion-parameter language backbone, 49 billion active parameters per text token and a 1,048,576-token context window, under a license with commercial conditions for model-as-a-service operators.
Microsoft launches seven models at Ignite; claims MAI Code 2 matches 96% of Claude Opus 5.5 at 5% of price
Microsoft unveiled seven AI models at Ignite on Nov. 17, aimed at coding, cybersecurity and other enterprise tasks. The company claims MAI Code 2 delivered 96% of Claude Opus 5.5's performance across four coding benchmarks at 5% of the price, using a promotional rate.
Comments
Loading...