# Is the best coding model better than a principal engineer? Not yet, but closer than the titles suggest (slim) > Cognition released FrontierCode 2.0 on 27 October 2026, 132 tasks written by open-source maintainers and graded on whether the pull request should be merged. Claude Opus 5.5 leads at 45.5%. Vlad Cealicu compares what it grades with what we expect of a staff or principal engineer, the top of the field for writing code, and estimates how one would score. - Full: https://www.anchorterminal.com/blog/frontiercode-principal-engineer.md (~3,400 tokens) · this version ~1,630 tokens · JSON https://www.anchorterminal.com/blog/frontiercode-principal-engineer.json · canonical https://www.anchorterminal.com/blog/frontiercode-principal-engineer - Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-10-28 - Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data) By Vlad Cealicu, 28 October 2026. 1,729 words, about 8 min to read. Cognition, the company behind the Devin coding agent, released FrontierCode 2.0 on Tuesday 27 October 2026 [[1]](#ref-1). I think this is a good thing to test. - 45.5%: Claude Opus 5.5's score, the best on the leaderboard - 132: tasks, each written by a maintainer of the repository it comes from - 40+: hours of a developer's time to build each task - 501: blocking checks for thoroughness, the biggest category, up from 168 FrontierCode 2.0 as published on 27 October 2026. Source [1]. ## What changed from the first version? FrontierCode 1.0 came out on 8 June 2026, built with the maintainers of 36 open-source repositories, with each task taking more than 40 hours of a maintainer's time [[3]](#ref-3). Version 2.0 keeps the method and makes the tasks harder. ## How is it graded? Each task has a rubric, and any check marked as blocking that fails scores the whole attempt zero [[1]](#ref-1). The tasks aren't public, to keep them out of training data [[3]](#ref-3). Score on FrontierCode 1.1 and 2.0, rounded, in per cent | Model | 1.1 | 2.0 | | --- | --- | --- | | Claude Opus 5.5 | 55 | 46 | | GPT-6 Astra | 53 | 43 | | Grok 4.7 | 48 | 29 | | Gemini 4 Argon | 48 | 26 | | Muse Wonder 1.0 | 47 | 24 | | SWE-2 | 50 | 24 | | DeepSeek V4.1 Flash | 44 | 19 | | Kimi K3 | 44 | 18 | Each series has a zero-based scale. Every model dropped, and the drop is uneven. ## What do we expect from a principal engineer? GitLab publishes its career framework, and for a principal engineer it asks for someone "Able to deliver results despite starting from unclear requirements", who "Makes responsible decisions and evaluates tradeoffs" and "Performs Code Reviews across the GitLab Codebase on a regular cadence" [[4]](#ref-4). Put those next to the code itself and I'd look for three things in a staff or principal engineer's pull request. FrontierCode grades all three. ## Two tasks that show the gap In the Mock Service Worker task, `resetHandlers()` didn't restart generator resolvers. In the PostHog Android SDK task, a race between the live queue and a legacy class, `PostHogSendCachedEventsIntegration.kt`, sent events out of order. That second one shows up across the whole benchmark. ## Would a principal engineer do better? Cognition doesn't publish a human score, so this part is my estimate, not a measurement. It would be hard for a principal engineer to jump into something they've never seen. Pass rate on FrontierCode 2.0, in per cent | Who | Pass rate | | --- | --- | | The repository's own maintainer | 95 | | A principal engineer, new to the repo | 55 | | A staff engineer, new to the repo | 50 | | Claude Opus 5.5 | 46 | | A senior engineer, new to the repo | 45 | | GPT-6 Astra | 43 | | A mid-level engineer, new to the repo | 25 | Each series has a zero-based scale. So are the models better than a staff or principal engineer? Not yet on this test, and the gap is smaller than the titles suggest, because a cold repository takes away most of what a principal's experience buys them. ## What's next? Cognition says it'll run the evaluation for every model maker, since the tasks stay private [[3]](#ref-3). We list [Devin](/tools/devin) on Anchor Terminal, and we grade the tools agents call, not how well a model writes code, so FrontierCode is a number I'll keep reading rather than one we compute. ## The terms, explained ### What is FrontierCode 2.0? Cognition's coding benchmark, released on 27 October 2026. It has 132 tasks written by maintainers of open-source repositories and grades whether a model's pull request is good enough to merge, using rubrics where any failed blocking check scores the attempt zero. ### Which model scored best? Claude Opus 5.5, at 45.5%, with GPT-6 Astra second at 42.8%. ### What does underspecified mean here? The prompts leave out engineering decisions on purpose, so the model has to work out what the maintainer wants from the codebase, such as fixing the same bug in a sibling code path or keeping the public API intact. ### Are AI models better than a principal engineer at coding? Not on FrontierCode 2.0, by my estimate. ### Why do some models score lower at higher reasoning effort? At higher effort they write more code and touch more files, so they fail more scope checks, such as deleting a file that's part of a public API, even though they get more thorough. --- More in [Opinion](https://www.anchorterminal.com/blog/opinion/index.md) · Older opinion [Genie 4 runs a Gemini agent alongside the world model](https://www.anchorterminal.com/blog/genie-4-world-model.md) - Older: [Can you run Qwen3.8-27B locally? Yes, on a graphics card you might already own](https://www.anchorterminal.com/blog/run-qwen3-8-27b-locally.md) - All posts: https://www.anchorterminal.com/blog/index.md ## References Read on 28 October 2026. Quotes are as published. [1] Cognition, "FrontierCode 2.0", 27 October 2026. Source for the release date, 132 tasks, 55% more repositories, security and performance tasks, 43% more blocking checks, 67% more files and 89% more lines, codebases 25% larger, 80% of reference solutions differing from upstream, the blocking checks by category, the thoroughness examples, scores at each effort level, the Mock Service Worker and PostHog tasks, and the 1.1 and 2.0 scores. [2] Cognition, "FrontierCode leaderboard". [3] Eric Lu, Ben Pan and others at Cognition, "Introducing FrontierCode", 8 June 2026. Source for 36 repositories, 40 hours per task, prompts a third the length of SWE-Bench Pro's, "expects the agent to infer the maintainer's intent", the grading methods, the QC quote, private tasks and the offer to model makers. [4] GitLab, "Dev Career Framework, Principal Engineer", GitLab Handbook. [5] Google, "The Standard of Code Review", Google Engineering Practices. [6] Cognition, announcement of FrontierCode 2.0, 27 October 2026. Source for "most coding evals" checking a model meets a spec, 1.0's "would you actually merge the PR?" and "not only would you merge this PR, but should you?".