AI Models
Flower Offers Endeavor Private Deployment, Not Public Weights
Flower Labs launched Endeavor 1.0 on September 1 as a generalist model available through a managed service or deployment inside a customer's infrastructure. The release does not announce public weights, and its benchmark table is company-reported; private deployment can improve control without proving model ownership, reproducibility, performance, or lower total cost.
Citation-ready: Flower Labs launched Endeavor 1.0 on September 1, 2026, with managed-service and private-deployment options, while the announcement did not release public model weights.
What happened and why it matters
No. Flower offers deployment inside customer infrastructure, but the launch does not publish weights or an open license, and the benchmark comparison still needs independent reproduction and workload testing.
Official Flower Labs model announcement
Primary reference: Flower Labs: Introducing Endeavor 1.0. Kaleido Field checked the event date and the article's attributed facts against this source.
| Source date | September 1, 2026 |
|---|---|
| Checked by Kaleido Field | September 2, 2026, 08:10 CST |
| Source function | current model-access analysis separating managed service, private deployment, model custody, public weights, API compatibility, company benchmark results, integration support, reproducibility, and total ownership cost |
Private deployment changes the control plane
Running a model inside an organization's environment can keep network paths, identity, logs, scaling, and sensitive workloads under local controls. It does not automatically transfer the artifact, training recipe, source code, or rights to modify and redistribute it.
A procurement record should name who holds weights, where they run, who can inspect or export them, license limits, update authority, telemetry, support access, key ownership, termination, and portability.
A four-benchmark table is a screening signal
Flower reports strong scores and includes competing model values. That is enough to justify reproduction, not to establish a universal frontier rank.
Teams should verify exact versions, prompts, tools, sample counts, pass rules, contamination controls, variance, and cost, then run their own tasks for accuracy, latency, safety, maintainability, and accepted-output cost.
Evidence boundary
Official product facts: model launch, managed and customer-environment deployment paths, familiar API formats, integration support, and private-deployment availability. Flower-reported evaluation: GPQA 92.0, HumanEval 98.2, AIME 2026 99.9, and IFEval 94.1 in the launch table. Not established: public weights, open license, independent benchmark reproduction, training-data disclosure, security review, service level, hardware requirements, total cost, or superiority on customer workloads.
FAQ
Can Endeavor run in a customer environment?
Flower says private deployment is available with integration support.
Are the weights public?
The announcement does not release public weights or an open license.
Who ran the benchmark comparison?
The scores are presented by Flower in its launch announcement.