Flower Labs announced Endeavor 1.1 on November 3, adding image understanding and an EU inference option to its reasoning and coding model for multi-step agent work. Flower also reports stronger reasoning and coding than version 1.0, based on its own evaluations. The details are in the official announcement.
Images, deployment choice and reusable work
The practical offering combines:
- Images alongside text: work with screenshots, charts and diagrams.
- Deployment choice: Flower-managed inference on global or EU GPUs, plus supported private-cloud or on-premises deployment.
- Reusable work: inspect readable thinking traces and retain outputs and traces, including for training custom models.
Access is by request. Model artifacts are privately distributed, not a public open-weight download, according to the developer documentation.
Why it matters
For app teams and agent builders, the interesting combination is visual input plus deployment control. It could support workflows that inspect a screenshot, investigate code and produce a reviewed fix while giving teams a choice about where inference runs.
The useful test is performance on their own tasks, not a headline benchmark ranking. Readable traces can help investigations, but do not prove an answer is correct.
For another approach to visual coding workflows, see our coverage of Meituan’s LongCat-2.5.