Free
$0
free forever
Try Aether without a payment method. Capped throughput, the smaller model variant, and the public benchmark cases.
- Aether 7B endpoint
- Limited monthly calls
- Public benchmark suite
- Community support
- No commercial deployment
Pricing
Honest pricing, at every scale.
One subscription covers over 100 products. There is nothing to buy per product and no quote to request — you pick a tier, and the tier decides how much you can run. Every rate below is published, which is more than any incumbent in simulation, EDA or drug discovery will tell you without a sales call.
Tiers differ by usage, not access. A five-second property lookup and an overnight GPU sweep cost us three orders of magnitude apart, so what you pay follows what you burn — measured in credits at a rate we publish, drawn first from the allowance your plan already includes.
swipe horizontally to see the full chart →
Pro
$20
/ month
For individual scientists and engineers running real workloads on their own. Higher limits, the standard model, full discipline access.
- Aether 40B endpoint
- Every discipline
- Streaming rollouts
- Higher monthly calls
- Email support
Max 5×
$100
/ month
Five times the Pro allowance, for engineers running sustained campaigns rather than occasional jobs.
- Everything in Pro, plus —
- 5x the Pro usage allowance
- Priority access at peak load
- Long-horizon session persistence
- Eval-harness access
Most popular
Max 20×
$400
/ month
For power users running long-horizon campaigns. Frontier model, expanded throughput, priority queue.
- Everything in Max 5×, plus —
- Aether 280B sparse-MoE
- 5× the monthly call limit
- Priority queue at peak load
- Long-horizon session persistence
- Eval-harness access
- API key + Python/TS SDKs
Team
$25
per seat / month · 2 seat minimum
For working teams. Shared workspace, role-based access, audit log, central billing. Annual contract.
- Everything in Pro, per seat
- Shared workspaces + handoffs
- Role-based access control
- SSO + SCIM provisioning
- Audit log
- Central billing + invoices
- Admin console
- Standard SLA
Enterprise
Custom
billed annually
For organisations replacing a stack of seven-figure renewals. Custom training, custom evals, custom safety.
- Everything in Team, plus —
- VPC + on-prem deployment
- Air-gapped option, ITAR / EAR-clean
- Custom fine-tunes + red-team
- Forward-deployed engineer
- 24×7 incident response
- Model-card co-authorship
- MSA + custom SLAs
Consumption rates
Tokens, compute, seats — three honest meters.
Aether is a simulation-native model, so it has two cost lines the chat-trained labs don't: solver compute and per-seat runtime. Token rates match what Frontier model providers charge for comparable model scales — the rest is priced the way AWS prices a GPU-hour.
Meter 01 · per million tokens
Tokens for reasoning
For reasoning, planning, dialog, code edits and agent decisions. Cached input cheaper. Batch API ships ~50% off.
| Variant | Best for | Input · $/1M | Cached input · $/1M | Output · $/1M |
|---|---|---|---|---|
| Aether Edge 1.3B | On-device · embedded | $0.20 | $0.05 | $0.80 |
| Aether 7B | Latency-sensitive · CI | $0.50 | $0.10 | $2.00 |
| Aether Bio 12B | Discovery · ADMET · FEP | $1.40 | $0.30 | $5.60 |
| Aether 40B | General workhorse | $3.00 | $0.60 | $15.00 |
| Aether 280B sparse-MoE | Frontier · long-rollout | $6.00 | $1.20 | $30.00 |
| — Frontier chat model · mid | comparison · chat | $3.00 | — | $15.00 |
| — Frontier chat model · large | comparison · chat | $2.00 | — | $8.00 |
| — Frontier chat model · top | comparison · chat | $15.00 | — | $75.00 |
Credits are the only currency. The token column is not a separate allowance — it is what the same credits buy if you spend the whole month on a mid-range general text model at $1.20 per million output tokens. A smaller model buys more, a frontier multimodal one fewer, and compute on Meter 02 draws from the same pool.
Comparison rows shown for reference at time of publication · not affiliated
Meter 02 · per compute-unit
Compute for simulation
Everything is metered in one credit, worth one cent, burned at a rate set by the hardware the work runs on. Five rates cover the whole catalogue — not one price per product, and not a second currency to reconcile against the first.
| Solver tool | Used by | Burn class | Credits / min | A typical run |
|---|---|---|---|---|
| CFD · transient | /v1/simulate | CPU HPC | 10 | 30 min, 48 cores ≈ 926 cr · $9.26 |
| FEA · nonlinear | /v1/simulate | CPU HPC | 10 | 20 min, 16 cores ≈ 333 cr · $3.33 |
| MD · classical, 1 ns | /v1/simulate | GPU heavy | 25 | 2 min ≈ 50 cr · $0.50 |
| DFT · periodic | /v1/simulate | CPU HPC | 10 | 45 min, 32 cores ≈ 1,143 cr · $11.43 |
| FEP · alchemical | /v1/fep | GPU heavy | 25 | 40 min ≈ 1,000 cr · $10.00 |
| ADMET · panel of 18 | /v1/admet | CPU light | 1 | 5 s ≈ 1 cr · $0.01 |
| Docking · induced-fit | /v1/dock | GPU mid | 10 | 30 s ≈ 5 cr · $0.05 |
| RTL synthesis · 7nm | /v1/rtl/synthesise | CPU HPC | 10 | 45 min, 16 cores ≈ 748 cr · $7.48 |
| Lab protocol planning | /v1/lab/plan | CPU light | 1 | 20 s ≈ 1 cr · $0.01 |
1 credit = $0.01 · multi-core work is charged on a tiered marginal rate, so 48 cores costs 3.1× four cores, not 12×
Meter 03 · per seat
Seat for the runtime
The IDE adapters, long-horizon agent runtime, memory store, capability-policy engine, audit log. Each seat ships with a bundled token + compute allowance; overage rolls into Meters 01 and 02 at standard rates.
| Tier | $ / seat / month | Included credits | Usage vs Pro | Equivalent tokens |
|---|---|---|---|---|
| Free | $0 | 500 credits | 0.2× | ≈ 4.2M output tokens |
| Pro | $20 | 2,500 credits | 1× | ≈ 21M output tokens |
| Max 5× | $100 | 12,500 credits | 5× | ≈ 104M output tokens |
| Max 20× | $400 | 50,000 credits | 20× | ≈ 417M output tokens |
| Team | $25 / seat | 2,500 credits / seat | 1.25× | ≈ 21M output tokens |
| Enterprise | Custom | Committed pool | Custom | — |
Cost calculator
Three worked examples. Real numbers.
Concrete cost estimates for three representative workloads — across the three meters. Numbers are typical; your real bill depends on workload size and reserved-capacity mix.
Calculator
Pick a workload. We do the math.
Numbers are typical and rounded. Your real bill depends on workload size and reserved-capacity mix.
Workload
Size · 4M cellssize × 1.00
Runs / month · 4
Estimated monthly bill
$559
per month at this configuration
- Meter 01 · tokens$4.99
- 0.32 M output · 0.06 M input
- Meter 02 · compute$154
- 48 CU at on-demand rate
- Meter 03 · seats$400
- 1 × Max 20× seat
Estimate for budgeting only — pilots, reserved capacity and enterprise agreements have different rate structures. Talk to sales for an exact quote.
One CFD design study
Transient blade · 18M cells · 12 s horizon
- Meter 01 · tokens
- $2.40
- Meter 02 · compute
- $11.50
- Meter 03 · seat
- $20/mo
~80k output tokens
~1,150 credits
1 Pro seat
≈ $13.90 per study + seat
FEP campaign · pharma
120 ligands × alchemical free energy on 4 targets
- Meter 01 · tokens
- $6.72
- Meter 02 · compute
- $1,150
- Meter 03 · seat
- $125/mo
~1.2M output tokens
~115,000 credits
5 Team seats
≈ $3.8k campaign + 5 seats
Whole-repo coding agent
200 PRs/month · review + tests
- Meter 01 · tokens
- $900
- Meter 02 · compute
- $0
- Meter 03 · seat
- $250/mo
~60M output tokens
Negligible · no solver
10 Team seats
≈ $1.2k/mo for the team
Side by side
Every feature, every tier.
The full matrix. Bookmark this page when you're filling out the procurement spreadsheet — we keep it accurate.
| Capability | Free | Pro | Max 5× | Max 20× | Team | Enterprise |
|---|---|---|---|---|---|---|
| Aether 7B | Yes | Yes | Yes | Yes | Yes | Yes |
| Aether 40B | — | Yes | Yes | Yes | Yes | Yes |
| Aether 280B sparse-MoE | — | — | — | Yes | Yes | Yes |
| Specialised checkpoints (Bio, Edge) | — | Bio only | Bio only | Yes | Yes | Yes |
| Monthly calls (typical) | 500 credits | 2,500 credits | 12,500 | 50,000 | 2,500 / seat, pooled | Committed pool |
| Streaming + steerable rollouts | Partial | Yes | Yes | Yes | Yes | Yes |
| Tool-use + memory | — | Yes | Yes | Yes | Yes | Yes |
| Eval harness | — | — | — | Yes | Yes | Yes |
| API key + SDKs | — | — | — | Yes | Yes | Yes |
| Shared workspaces | — | — | — | — | Yes | Yes |
| SSO + SCIM | — | — | — | — | Yes | Yes |
| Audit log | — | — | — | Yes | Yes | Yes |
| VPC deployment | — | — | — | — | Add-on | Yes |
| On-prem / air-gapped | — | — | — | — | — | Yes |
| GxP / 21 CFR Part 11 ready | — | — | — | — | Add-on | Yes |
| Custom fine-tunes | — | — | — | Limited | On request | Yes |
| Forward-deployed engineer | — | — | — | — | — | Yes |
| Response SLA | Community | 1 business day | 4 business hours | 24×7 |
What this looks like at scale
From solo to seven-figure deal.
Four sample shapes spanning the full pricing range. Most customers move up a tier as their workload deepens.
Solo researcher
$20 / month
1 user · Pro
A researcher running occasional simulation or chemistry work. Pro covers every product and 2,500 credits a month — about $25 of compute, included.
Power user
$400 / month
1 user · Max 20×
Sustained campaign use — long-horizon agent sessions, frontier 280B-MoE checkpoint, eval harness, API key with Python and TypeScript SDKs.
Working team
$2,880 / year
12 seats · Team
Shared workspaces, SSO, audit log, central billing. Annual commitment. Covers a small discovery, CAE, or chip-design team running production workloads together.
Mid-sized aerospace prime
Custom
Enterprise
Replaces a $4.6M renewal across structural, fluid, EM and crash. Pilot ran 7 weeks; pilot fee was credited to year-one MSA.
Questions we get
The things people ask before signing.
If your question is not here, write to hello@apexworldlabs.com. A real engineer answers, not a sequence of canned auto-replies.
How is this priced compared to other AI assistants?
Same shape — Free, Pro ($20/mo), Max 5× ($100/mo), Max 20× ($400/mo), Team ($25/seat/mo), Enterprise (custom). One subscription covers every product; the tiers differ by how much you can run, not by what you can open.
Do you train on our data?
No. Base models are never trained on customer prompts, code or experiment data. Fine-tunes are scoped to the customer that requested them and never re-used.
Can I switch tiers mid-month?
Yes. Upgrades take effect immediately and we prorate the difference. Downgrades take effect at the next billing cycle.
Is Pro enough for real work?
For an individual scientist running occasional simulation or chemistry workloads, yes. For sustained campaigns, sustained CFD runs, or chip-design flows, you will want Pro.
What counts as a "call"?
One API request to Aether. A streaming trajectory counts as one call regardless of length. Tool calls are not counted separately. We publish exact limits per tier in the customer console.
Can we run it on-prem?
On-prem and air-gapped deployment is Enterprise-tier only. VPC deployment is available as a Team add-on; the Enterprise tier includes it.
Do you have a SaaS endpoint?
Yes, on every paid tier. Pro and Max hit our managed cloud; Team can route through VPC; Enterprise can run sovereign.
What if we already have commercial CAE / EDA / comp-chem tools?
Most enterprise customers do, at the start. We run a paid pilot on one workload — typically scoped at the Team or Enterprise tier — you compare the numbers, and you decide what stays.
How are seats counted?
By unique humans authenticated against the platform in a given month. Read-only viewers do not consume seats. Service accounts attached to CI or to instruments are bundled rather than counted as seats.
Is there an academic discount?
The Free tier covers most academic non-commercial use. For research groups needing more, write to research@apexworldlabs.com — we have free Pro-tier access for academic safety research and many published-research workloads.
Above Enterprise
Frontier — for organisations building at the edge.
Paired senior research engineer, pre-release roadmap access, dedicated training compute, joint research output. Selective — six to ten global engagements at any time.
Start free. Scale when ready.
Most people start on Free, move to Pro for sustained use, step up to Max when they outgrow the allowance, and graduate to Team or Enterprise when their organisation joins. There is no penalty for starting small.