Endeavor 1.0

Endeavor 1.0 is Flower Labs’ new generalist model for reasoning, coding and long-horizon agent work. It is designed to preserve context across many steps, use tools, recover from failures and work through familiar model APIs and response formats. ¹²

Flower offers two deployment paths: a managed service operated by Flower and private deployment inside infrastructure controlled by the customer. The latter is the product’s clearest differentiator for organizations that want frontier-class capability without making every workload dependent on a closed remote API. ¹²

Flower reports very strong public-benchmark results, including 92.0 on GPQA, 99.9 on AIME 2026, 98.2 on HumanEval and 94.1 on IFEval. AiToolMap treats those numbers strictly as provider-reported claims: the current fixed panel did not produce an independent exact-model benchmark result that could validate or normalize them. ¹

Endeavor launched on September 1, 2026 as a preview rather than a self-service public model. Access is by request while Flower onboards a limited group of organizations and partners and expands compute availability. ¹

No stable public monetary price was verified on the current launch, model or legal surfaces. AiToolMap therefore does not invent token prices, seat prices or private-deployment fees. Buyers should expect a sales-led enterprise process and verify the commercial structure for managed versus private deployment directly with Flower. ¹²

The lack of public pricing materially limits value-for-money comparison. Private deployment may create strategic value for organizations with sovereignty, data-location or infrastructure-control requirements, but the current public evidence does not let AiToolMap quantify that benefit against contract and operating cost.

The strongest evidence for Endeavor’s present existence and deployment model is first-party. Flower’s launch post and model page are explicit about preview access, managed service, private deployment, API compatibility and the intended reasoning/coding/agent scope. Same-day Tech.eu coverage independently corroborates the launch and the managed/private deployment positioning, but does not turn Flower’s benchmark table into independent performance evidence. ¹²³

The fresh fixed 50-source panel found no attributable current exact-product numeric rating family. Seven panel sources were technically inaccessible in this cycle; the other targeted source-domain searches returned no exact Endeavor 1.0 result. Missing evidence is not scored as zero.

Flower has useful organization-level security evidence. The company states that it is ISO/IEC 27001 certified, with externally reviewed controls covering areas such as access control, encryption, operations, communications and vendor management. Its current legal index also exposes current Terms, Privacy Notice, DPA, AUP and Trust Center. These are positive governance signals, not a product-specific promise that every Endeavor deployment has one universal retention or data-flow model. ⁴⁵

Private deployment changes the trust boundary materially because Flower says the model and workloads can remain in the customer’s own environment. Buyers still need to validate the exact architecture, support access, telemetry, update path, logging, model/data retention and incident-response responsibilities for their chosen deployment. ²

Endeavor is aimed primarily at organizations building substantial coding, reasoning or autonomous-agent workloads that care about where the model runs and want the option to move from a managed service toward infrastructure they control. It is particularly relevant to teams that see model portability and deployment sovereignty as strategic requirements. ¹²

It is less suitable today for individual buyers or teams that need immediate self-service access, transparent token pricing, broad third-party integrations or a long public operating history. The preview model means prospective users should treat procurement, capacity and support as part of the evaluation rather than assuming consumer-style availability.

A serious evaluation should reproduce Flower’s claims on the buyer’s own repositories, tool chains and long-horizon tasks, and should compare managed and private deployment operationally rather than deciding from provider benchmark tables alone.