Pay for compute,
not for lock-in.
Transparent pricing across plans. Every tier includes your weights, your data, and no surprise egress fees.
Starter
For experiments and proofs of concept on our shared cluster.
- Fine-tune Axon Mini (7B)
- $2.40 per GPU-hour, pay as you go
- 10k inference calls / month included
- Community support
- Public eval dashboards
Scale
For teams putting custom models into production.
- Everything in Starter
- Fine-tune Axon Pro (70B)
- $1.90 per GPU-hour (20% off)
- 2M inference calls / month included
- Private VPC deployment
- Eval gates + drift monitoring
- Priority support, 4-hour SLA
Enterprise
For organizations training bespoke frontier models.
- Everything in Scale
- Dedicated GPU reservations
- Pretrain from scratch (Axon Custom)
- On-prem / air-gapped deployment
- Dedicated research team
- SOC 2, HIPAA, custom DPAs
- 24/7 support, 1-hour SLA
What's in each plan.
| Feature | Starter | Scale | Enterprise |
|---|---|---|---|
| Base models | Mini | Mini, Pro | Mini, Pro, Custom |
| GPU-hour rate | $2.40 | $1.90 | Reserved pricing |
| Included inference | 10k / mo | 2M / mo | Custom |
| Private deployment | — | VPC | VPC / on-prem / air-gapped |
| Eval gates | — | ✓ | ✓ |
| Own your weights | ✓ | ✓ | ✓ |
| Support SLA | Community | 4-hour | 1-hour, 24/7 |
| Compliance | SOC 2 | SOC 2 | SOC 2, HIPAA, custom |
Questions, answered.
Yes. On every plan, the weights you train are yours. You can export them, deploy them anywhere, and keep them if you leave. We never hold your model hostage.
You're billed per second of actual GPU runtime, rounded to the nearest minute, at your plan's rate. Idle time and queueing are never charged. You can set hard spend caps per project.
Absolutely — that's the point. Upload your corpus and it stays in your isolated environment. On Scale and Enterprise it never leaves your VPC. We run provenance and license checks but never train shared models on your data.
Overage is billed at standard per-call rates, which decrease with volume. You'll get alerts at 80% and 100% of your included usage so there are no surprises.
Yes. We offer credits for early-stage startups and academic research groups. Get in touch and tell us about your project.
Still deciding?
Talk to our team and we'll help you scope the right plan for your workload.