World Model Optimizer Pricing
Verified:
No public Experiential Labs price is available. World Model Optimizer 0.2.0 is published on GitHub and PyPI, but there is no public hosted-platform rate card, subscription price, or universal WMO token rate. Model inference, training, evaluation, and sandbox usage can still create substantial third-party charges.
World Model Optimizer price status
| Product layer | Published price | Availability |
|---|---|---|
| WMO Python package | No package fee listed | Version 0.2.0 on PyPI |
| Hosted platform | Not public | Sign-in UI online; launch status unclear |
| Model inference | Provider rates | Bring registered model providers |
| Distillation and evaluation | Provider rates plus infrastructure | Tinker, E2B, and other backends supported |
What the 40% saving means
Experiential Labs says WMO can start with a frontier-quality endpoint at 40%+ lower cost. Its official routing benchmark supports that narrower wording: the cost-saver anchor measured 40.8% lower cost with a 0.87-point quality gain, while the maximum-savings anchor measured 46.2% lower cost with a 0.54-point quality loss.
These are routing outcomes against the best single model in one held-out nine-model cohort. They are not discounts on a WMO rate card. The documentation says the cohort is favorable to routing because its models score within a few points of one another and tells users to re-measure on their own traffic.
Training can cost more than the package
WMO's Qwen reference distillation run authorizes up to $1,200 and estimates roughly $1,024 at cold teacher rates. The bill comes from student rollouts, teacher scoring, training tokens, evaluations, and the task environment. A cheaper routed endpoint must save enough production spend to recover those setup costs.
For the benchmark numbers, distillation maturity, and model-price context, read our World Model Optimizer cost analysis. You can also compare current provider token rates in the AI API pricing table.
Canonical pricing and Labs status
WMO is not included in our pricing API because it is not a separately priced model endpoint. Adding an estimated or zero-dollar record would blur the package cost with the third-party models it routes.
It is also unavailable for the current cost-per-task leaderboard. Labs needs a stable callable model ID, a public endpoint, and official rates. WMO produces a workload-specific router instead. We will review inclusion when Experiential Labs publishes a testable hosted endpoint or a complete self-hosted benchmark contract.
Frequently asked questions
How much does World Model Optimizer cost?
Experiential Labs has not published a package price, hosted-platform subscription, or universal per-token WMO rate. Users still pay the model, training, evaluation, and sandbox providers used by their setup.
Does World Model Optimizer cut AI cost by 50%?
Not as a guaranteed discount. Its official held-out routing benchmark measured 40.8% lower cost at the cost-saver setting and 46.2% at maximum savings. Results depend on the workload and model pool.
Is World Model Optimizer a model API?
No. WMO is an optimization, routing, distillation, and serving layer. It can expose a local OpenAI-compatible endpoint, but it does not publish one universal hosted model ID or token price.
Is World Model Optimizer open source?
Experiential Labs calls it open source and publishes the code on GitHub, but the repository did not expose a license file during our July 27 verification. Review the licensing terms before commercial use.