LongCat 2.5 review navigator
Official benchmarks, independent analysis, and community reports about LongCat 2.5, clearly separated from Tabbit's own testing.
Official
1 source-checked resourcesMedia
2 source-checked resourcesDEV: Real-Project Experience with LongCat 2.5 Preview and Space Bunny Alpha
The author says that across several of their own development projects, LongCat 2.5 Preview outperformed Space Bunny Alpha in areas such as context understanding, tool use, and handling blockers. This is a subjective account, and the article provides no quantif.
AI Coding Daily: LongCat-2.5-Preview Scores 44.65 in OpenCode Coding Tests
As of the collection date, the LongCat-2.5-Preview and OpenCode combination scored 44.65/70 on the site's coding leaderboard, ranking 44th. This is a snapshot of a specific task and tool combination.
Community
8 source-checked resourcesA Reddit User Tests LongCat 2.5 Preview on C# Code Analysis
The author says LongCat 2.5 Preview repeatedly looped while analyzing five C classes in OpenCode and still had not completed the request after about 20 minutes. This was one personal experience, without complete input or repeated tests.
Reddit: LongCat 2.5 Preview Compared with Two Other Models on TB 2.1
The author reports that LongCat 2.5 Preview was run three times on each of 89 TB 2.1 tasks, achieving a 48.7% Pass@1 rate and passing 64.0% of tasks at least once across three runs. This was an independent comparison with incomplete configuration details and r.
Reddit: LongCat 2.5 Preview Free Service Experiences Timeouts and 429 Errors
The author says the free LongCat 2.5 Preview service repeatedly timed out or lagged, and returned an HTTP 429 error stating that the upstream endpoint was unavailable. This is a personal report without configuration details or quantitative records.
Reddit: LongCat 2.5 Preview on a Large Refactoring Task
The author says that during a large refactor, LongCat 2.5 Preview repeatedly failed to use tools, asked the same questions again, and changed unrelated code, leaving a result that would not compile. This is a personal report without code or execution logs.
Inferwerk: Independent Test of LongCat 2.5 on EvalMapBench
Inferwerk reports that LongCat 2.5 (max) met 79.6% of detail checks in its EvalMapBench test on 2026-11-02; 51 of 129 trials met every check. The result reflects performance only on this private test set.
A Reddit User's Experience Porting a Three.js Project
In a large JavaScript/Three.js-to-Swift/Metal porting task, one user said LongCat-2.5-Preview got stuck in a loop, but resumed work after the user interrupted it and pointed this out. After further testing, the author personally felt that Space Bunny performed.
Elias: Front-End Experience with LongCat 2.5 Preview in OpenCode
The author says LongCat 2.5 Preview felt fast and concise when tried in OpenCode, with front-end results somewhat close to Muse spark 1.3. He also believes it still falls short of frontier models. This is a subjective personal assessment without reproducible c.
Reddit: LongCat 2.5 Preview First-Token Latency and Throughput Test
In a brief test, the author reported a 2.19-second time to first token and a generation speed of 30.0 tokens per second for longcat-2.5-preview-free; the sample and configuration are insufficient to represent long-term service performance.
LongCat 2.5
Use and compare models in Tabbit
Official benchmarks, independent analysis, and community reports about LongCat 2.5, clearly separated from Tabbit's own testing.