DOE Announces Genesis-Science-1: Pricing Impact
DOE and Arcee announced Genesis-Science-1, an open-weight model for scientific workflows. See its release status and cost implications.
By AI Pricing Guru Editorial Team
AI Pricing Guru articles are maintained by the editorial workflow behind the site: daily pricing snapshots, provider source checks, and review passes for model launches, subscription limits, and billing changes.
TL;DR
- The U.S. Department of Energy and Arcee announced Genesis-Science-1, the first planned model in DOE's Genesis Open Models Initiative.
- This is a development program, not a downloadable checkpoint or priced API launch: DOE has not published weights, a license, parameters, or a release date.
- GS1 is being designed for complete scientific workflows, including code modernization, simulation, materials science, and energy systems—not just question answering.
- Research teams should contribute or follow the project now, but keep current models in production until GS1 has release terms, benchmarks, and measured infrastructure requirements.
Cost comparison from today's pricing data
USD per 1M tokens. Input and output rates are charted separately.
Published managed-API baselines while GS1 pricing is unavailable
| Model | Provider | Input / 1M | Cached / 1M | Output / 1M |
|---|---|---|---|---|
| Mistral Large 3 | mistral | $0.5 | n/a | $1.50 |
| Qwen3 235B A22B Instruct 2507 | together | $0.2 | n/a | $0.6 |
| DeepSeek V4 Pro | together | $1.74 | $0.2 | $3.48 |
Built from pricing.json at publish time.
The U.S. Department of Energy announced the Genesis Open Models Initiative on August 7 and named Genesis-Science-1 (GS1) as its first planned open-weight foundation model. Arcee AI is the first industry partner.
The pricing takeaway is not a new token rate. DOE has opened contribution rounds for the model’s training and evaluation, but it has not released GS1 weights, a license, parameter count, hosted endpoint, or API price. The live table above shows current managed open-model baselines from our dataset; none is a proxy for GS1 pricing.
What DOE announced
GS1 is intended to work inside scientific environments that contain code, datasets, tools, documentation, logs, partial results, and failure states. DOE lists high-performance-computing code modernization, experimental analysis, simulation campaigns, materials science, and energy systems among the initial areas.
The planned workbenches may include Python, Fortran, C and C++, MPI, OpenMP, CUDA, HIP, notebooks, command-line tools, simulation packages, and computing schedulers. DOE says approved tools will run in governed, sandboxed environments with checkpoints, retry handling, and records of prompts, tool calls, code changes, data, intermediate artifacts, and conclusions.
That makes GS1 closer to a scientific workflow agent than a conventional chat model. Human review will remain required for decisions involving safety, security, publication, and resource use, and DOE says the model will not receive blanket access to its systems.
Release and access status
| Item | Status on August 8 | Cost implication |
|---|---|---|
| Model weights | Not released | No self-hosting benchmark can be verified yet |
| Public license | Not announced | Commercial and redistribution rights remain unknown |
| Hosted API | Not announced | There is no official token rate or service-level commitment |
| Model size and architecture | Not disclosed | GPU memory, throughput, and cluster requirements cannot be budgeted |
| Contribution program | Open | Selected contributors receive early evaluation access, not a public production endpoint |
The first foundation-stage contribution window closes August 14, with selected material due August 28. The first post-training data and environment window closes August 25, with delivery due September 14. DOE says further deadlines are expected every three months.
Pricing impact
Open weights could lower vendor lock-in and let laboratories deploy on infrastructure they control. They do not make inference free. Scientific agents can consume substantial GPU time through long tool-driven runs, simulations, retries, code execution, storage, and expert review—even when the model license itself has no fee.
GS1’s focus on reproducible workflows may improve total economics if it reduces failed experiments, abandoned simulation runs, or manual code modernization. That benefit cannot be priced from the announcement. Buyers will need task-completion rates, wall-clock time, accelerator-hours, retry counts, and expert-review time from a pinned release.
Arcee’s prior Trinity program culminated in a sparse mixture-of-experts model with 400 billion total parameters, but DOE has not said GS1 will reuse that architecture or footprint. Do not use Trinity hardware estimates as a GS1 budget.
What research teams should do
- Apply to the relevant DOE contribution track if you can supply scientific data, complete workflow environments, evaluations, or expert review under clear usage terms.
- Keep GS1 out of procurement estimates until DOE publishes a checkpoint, license, architecture, evaluation report, and deployment guidance.
- Build a production-derived test set now, including failed runs and recovery cases—not only clean question-answer tasks.
- Compare cost per reproducible result, including compute, tool execution, storage, and human review.
- Retain a managed fallback while testing any future self-hosted GS1 release.
For current alternatives, compare Mistral pricing, Together AI pricing, and DeepSeek pricing, then model token workloads in the calculator. Our open-model hosting comparison explains why model access and production economics are separate decisions.
Teams that want to prototype on managed open models can review Novita’s current catalog. Verify the exact model and version before treating any listing as a GS1 route.
Affiliate disclosure: AI Pricing Guru may earn a commission from the sponsored Novita link at no extra cost to you. It does not affect this analysis.
Bottom line
Genesis-Science-1 is an important public-sector open-model program, but it is not yet a released or priceable model. Its strongest promise is workflow-level scientific reasoning with auditable execution—not a confirmed token discount.
Follow the development program and prepare a realistic evaluation set now. Make deployment or budget decisions only after DOE publishes the weights, terms, benchmarks, and infrastructure requirements.
Sources: the U.S. Department of Energy’s official Genesis Open Models Initiative announcement, the initiative’s program overview, and the official Genesis-Science-1 description. Official sources and pricing status checked August 8, 2026.