Starting today, builds in Lovable are faster, without any drop in quality. Opus 5.5 is now available in Lovable, and it gets you the same results as Opus 5 while finishing in a third to half fewer steps. That means less waiting on long edits, and no compromise on what gets shipped.
Under the hood, that's because Opus 5.5 gathers context in one pass, makes fewer and more complete edits, and doesn't circle back unnecessarily to re-check its work.
It's also more disciplined about verification. Our verification-discipline benchmark measures whether a model checks a change as much as the task needs and then stops, rather than re-opening the browser or repeating the same check. Opus 5.5 scored higher than Opus 5 at every effort level while taking roughly half the steps.
This post breaks down how we benchmarked it, where it wins, and what it means for your builds.
How we evaluate new models
We run every candidate model against Lovable's internal benchmark suite, which measures end-to-end app-building performance on:
- 0-to-1 building: building from scratch with Lovable and achieving the desired result for our users
- Fixing and evolving existing codebases: going back into a live app to add a feature, fix a bug, or rework a screen without breaking what's already there (the largest category of work builders do in Lovable)
- Verification discipline: checking a change as much as the task needs and no more, then finishing inside the budget instead of over-verifying, repeating the same check, or looping on failed attempts
We evaluated Opus 5.5 against Opus 5 at equivalent reasoning effort levels: low, medium, and high, on the same code and with the same judges. Each task was run at least three times.
Opus 5.5 scored 4 to 6% ahead of Opus 5 on verification discipline. On 0-to-1 building and iterative code fixing the two models scored level. The tables below show where the efficiency gains are, in the number of individual steps each task took and in the tokens it processed along the way.
Individual steps per task deltas for Opus 5.5 vs. Opus 5
| Benchmark / Effort Level | Low | Medium | High |
|---|---|---|---|
| 0-to-1 building | -48% | -26% | -26% |
| Iterative code fixing | -47% | -34% | -34% |
| Verification discipline | -57% | -46% | -42% |
Input tokens processed per task deltas for Opus 5.5 vs. Opus 5
| Benchmark / Effort Level | Low | Medium | High |
|---|---|---|---|
| 0-to-1 building | -46% | -21% | -26% |
| Iterative code fixing | -36% | -27% | -29% |
| Verification discipline | -59% | -46% | -41% |
Every difference in both tables is statistically significant at the 95% confidence level. Input tokens are the context a model reads and re-reads across a task, so fewer of them is the direct footprint of gathering context once. Output tokens fell by 37 to 64% on iterative code fixing and verification discipline, and stayed close to Opus 5's on 0-to-1 building, where the work is generating the app itself.
What changed under the hood
Three things stand out in how Opus 5.5 handles Lovable workloads:
- It gathers context once. Opus 5.5 collects what it needs in a single pass, chaining file listings, searches and reads into one command, and it doesn't go back to re-inspect files between edits. On existing codebases this alone accounts for most of the steps it saves.
- Fewer, more complete edits. It patches several files in one round rather than editing, re-reading and editing again, and its changes are tighter: smaller diffs touching fewer files, for the same result.
- It verifies in proportion. Where Opus 5 could spend a dozen turns re-running the same check, Opus 5.5 checks what the change needs and finishes with far fewer output tokens, which is why verification discipline is the one benchmark where its performance leads.
What this means for builders
Opus 5.5 is one of the models Lovable now reaches for to give you the quality of Opus 5 in a fraction of the time. The gains come from how it works, not from cutting corners: one pass to gather context, fewer and tighter edits, and less time needlessly re-checking its work.
For Lovable builders, Opus 5.5 means faster builds with the same quality, whether you're starting from scratch or working on a live app. It gathers context once, makes fewer and more complete edits, and doesn't get stuck retrying, finishing in a third to half fewer steps and using significantly fewer tokens along the way.
— Fabian Hedin, CTO & Co-founder, Lovable
