Know what the frontier costs before you commit.
New frontier models reshape your cost and effectiveness math the moment they land. Frontier Preview Mode lets Cromus track and price preview and restricted models like Anthropic's Claude Mythos 5.1 as provisional, governed entries, so you can weigh the trade-off before they reach general availability — and without ever risking a deploy. It's exactly how Cromus tracked OpenAI's GPT-5.6 Sol, Terra, and Luna before they graduated to GA.
The frontier moves faster than your forecast
A new frontier model lands every few weeks, and each one quietly redraws your cost curve. The question every team asks is the same: if I move this workflow to the new model, what happens to my bill, and is the quality jump worth it?
The usual way to find out is to wait for general availability, wire the model in, run real traffic, and read the invoice a month later. By then the tokens are spent and the decision is already made. Frontier Preview Mode lets you reason about that trade-off up front — while a model is still in preview.
What Frontier Preview Mode does
Provisional pricing from day one
The moment a frontier model is announced, Cromus carries it with indicative preview pricing. You can see the cost shape of a restricted model like Claude Mythos 5.1 on a workflow while access is still limited — exactly as Cromus tracked GPT-5.6 Sol, Terra, and Luna before they graduated to GA.
Weigh cost against effectiveness
Frontier quality isn't free. Cromus puts a premium-tier frontier model and the balanced GA pick side by side on the same workflow, so you read the cost-per-quality delta — not just the headline benchmark.
Deploy-safe by design
A preview model is never auto-selected as the cheapest option. When one is pinned to a node, Cromus flags it as provisional and blocks deploy on that workflow until a generally available model is swapped back in.
The frontier models Cromus is tracking
Pricing is per 1M tokens and provisional — it mirrors each provider's published preview rates and updates as they firm up. Claude Mythos 5.1 is restricted to Anthropic trusted-access programs for cybersecurity and life sciences, not generally callable through the public API; it is tracked for modeling and non-deployable in Cromus. The GPT-5.6 family (Sol, Terra, Luna) and Claude Fable 5 have since graduated to general availability — they now price live from the Cromus model registry and have left this preview lineup. Cromus reflects each model's real status rather than pretending it's ready to ship.
Claude Mythos 5.1
Anthropic
Trusted-access frontier tier — the same underlying model as Claude Fable 5.1, with safeguards designed for advanced cybersecurity and life-sciences research. Available only to vetted Anthropic programs; not generally callable through the public API. Pricing preserved for modeling; not deployable in Cromus.
Indicative preview pricing for modeling purposes only. Final rates and availability are set by the model provider at general availability. Limited-access models are not deployable in Cromus; their pricing is preserved for modeling until access firms up.
Recently graduated to general availability
GPT-5.6 Sol, Terra, and Luna reached general availability on July 13, 2026, and Claude Fable 5 on July 1, 2026 (export controls lifted). Each has left the preview lineup above — Cromus now prices them live from the model registry and they are deployable in Cromus where your workspace scope allows. Tracking them provisionally before GA is exactly what Frontier Preview Mode is for.
Model a frontier upgrade
Pick a frontier model, choose a generally available model to compare against, and dial in your monthly run volume. Cromus prices both on the same workflow so you can read the cost-and-effectiveness trade-off before you commit a single token.
Frontier model
Trusted-access Anthropic frontier tier — the same underlying model as Fable 5.1, with safeguards designed for advanced cybersecurity and life-sciences research. Available only to vetted programs; not generally callable or deployable in Cromus.
Compare against
Runs per month
15,000Modeled at 6,000 input + 1,500 output tokens per run.
Claude Sonnet 4.6
Per run
$0.0405
Per month
$607.50
Claude Mythos 5.1
Per run
$0.1350
Per month
$2,025.00
Moving this workflow to Claude Mythos 5.1 would
add $1,417.50/mo (+233% vs Claude Sonnet 4.6)
Provisional model — frontier pricing is indicative preview pricing subject to change at general availability, and GA-baseline pricing is pulled live from the Cromus model registry. This calculator is for planning only; preview and limited-access frontier models are not currently callable in Cromus.
How Cromus handles a preview model
Frontier Preview Mode is a modeling layer, not a deploy path. A preview model only enters a simulation when it's explicitly pinned to a node — it is never auto-selected as the cheapest option, so a not-yet-shippable model can never quietly become your default.
Every node priced on a preview model is flagged as provisional, and any workflow that contains one carries a blocked-deploy banner until you swap back to a generally available model. You get the full cost-and-effectiveness picture for planning, with a hard guardrail against shipping something that isn't ready.
Frontier Preview Mode is rolling out on the Founding plan and above.
Model a frontier upgrade before you commit
Score a workflow, weigh a frontier model against your GA pick, and watch the cost and quality move — all before a single token is spent.