Aleph Alpha pricing

About

Aleph Alpha is a Heidelberg-based AI company — once positioned as Europe’s flagship answer to OpenAI — that today sells sovereign enterprise AI to corporations and the public sector rather than a public foundation-model API. Its current product center of gravity is PhariaAI, described as an “operating system for generative AI”: a platform that layers retrieval, agents, and workflows over multiple foundation models (including Aleph Alpha’s own small Pharia-1 family and third-party models), deployable on-prem or in a sovereign cloud so that companies and governments retain full control of their data.

Founded in 2019 by Jonas Andrulis and Samuel Weinbach, Aleph Alpha raised roughly €500M in November 2023 (led by the Innovation Park AI / Schwarz Group ecosystem, with Bosch, SAP and HP among backers) — at the time one of Europe’s largest AI rounds. But by September 2024 the company made an unusually candid strategic admission: it would step back from competing on raw frontier-model capability. CEO Jonas Andrulis put it plainly: “Just having a European LLM is not sufficient as a business model. It doesn’t justify the investment.” Lacking the compute capital of OpenAI or Mistral AI, Aleph Alpha repositioned around a deployable platform plus professional services, reportedly settling at around €20M ARR with roughly 200 staff and the Baden-Württemberg state government as an anchor public-sector customer.

The pricing consequence is the subject of this page: Aleph Alpha retired its public price card entirely. The legacy Luminous per-token API — once billed against prepaid EUR credits at premium rates — is wound down for new customers, and the former /pricing URL now returns a 404 (verified 2026-06-11). Pricing is sales-only: PhariaAI platform licenses, sovereign deployments, and services are quoted on contact. In April 2026 the story took a further turn — Cohere agreed to acquire/merge with Aleph Alpha in a roughly $20bn combined-valuation sovereign-AI deal — but as of this writing the public pricing posture remains contact-driven.

Pricing summary : how Aleph Alpha’s pricing model works today

Aleph Alpha’s pricing is sales-only and contract-based — a commitment / enterprise-license model, not a self-serve per-token API or a published subscription. There is no public price card to capture; the dimensions a buyer negotiates are:

• PhariaAI platform license — access to the sovereign-AI “operating system” (multi-model orchestration, RAG, agents, workflows), quoted per engagement. Typically an annual commitment.
• Deployment scope — on-prem or sovereign-cloud install, sized to the environment, data-residency requirements, and scale. This is where the European-sovereignty positioning becomes a price-able dimension.
• Specialized / domain models — training or fine-tuning smaller models (e.g. the Pharia-1 family) on the customer’s own data, sold as part of the platform or a services engagement.
• Professional services — discovery, build, integration, and change management, billed time-and-materials or per project. After the pivot, services revenue is a structural part of the model rather than an add-on.
• Legacy Luminous credits — existing customers may still draw down prepaid EUR credits per token on the wound-down Luminous API, but this is not offered to new buyers.

What makes this different: Aleph Alpha is a foundation-model lab that deliberately walked away from public per-token pricing. Where peers like Mistral publish raw token rates, Aleph Alpha converted its entire commercial surface into a quoted, sovereignty-led enterprise sale — the clearest example in this corpus of a lab retreating from the pure-usage API model into commitment-based platform-and-services pricing.

Pricing by product

Current offering (sales-only — all quoted on contact)

Legacy Luminous API (wound down — historical reference only)

USD-equivalents catalogued by third-party trackers; originally EUR-denominated and billed against prepaid credits drawn down per token. Not offered to new customers.

Sales motions across products: fully sales-led and partner-led. There is no self-serve or PLG motion — every current surface routes through a sales conversation, and the company emphasizes accompanying partners “from the first idea through implementation.”

Hidden costs : what Aleph Alpha buyers actually pay

Because pricing is fully quoted, the “hidden costs” here are structural rather than line-item: there is no headline rate to anchor against, so the real cost is assembled in a sales conversation. Two archetypes illustrate where the money goes.

Archetype 1 — a public-sector body adopting PhariaAI on-prem. A state ministry deploying PhariaAI for document analysis and file management (the Baden-Württemberg pattern). The platform license is one line, but the on-prem deployment, model specialization on internal documents, and the multi-month services engagement to integrate and roll it out typically dominate the total.

The lesson: for a sovereign deployment, services and integration usually outweigh the software license — you are buying an outcome (a compliant, data-resident AI system) assembled by Aleph Alpha’s team, not a metered API you wire up yourself.

Archetype 2 — a legacy Luminous API user. A customer still on the wound-down API draws down prepaid EUR credits per token. Here the surprise is the premium: Luminous Supreme at roughly $175 in / $192.50 out per 1M tokens is far above commodity model rates, so an output-heavy workload accumulates cost quickly — and the product is no longer being invested in, which is its own migration risk.

Want to estimate your own Aleph Alpha bill? Use the Aleph Alpha pricing calculator to model historical Luminous token costs — note that current PhariaAI pricing is quote-only and cannot be modeled from a public rate.

Pricing evolution : Aleph Alpha pricing history and changes

Aleph Alpha’s pricing arc is one of the most instructive lifecycle stories in this corpus: a frontier-lab that started with a public, premium per-token API and then deliberately dismantled its own price card in favor of a quoted, sovereignty-led enterprise model. The dated milestones below are reconstructed from primary press, the company’s own statements, and third-party price catalogues.

Cadence

Tracked range: 2023 Q2–2026 Q2. The defining change is not a price move but a pricing-model retreat — from public per-token to fully quoted enterprise.

Notable changes

• 2023-04 — Luminous API live: Base, Extended, Supreme billed per token against prepaid EUR credits; Supreme priced at a premium (~$175 in / $192.50 out per 1M tokens).
• 2023-11 — ~€500M Series B (Schwarz Group ecosystem, Bosch, SAP, HP) funds the frontier ambition.
• 2024-08 — Pharia-1-LLM-7B open-weighted; product gravity shifts toward deployable models.
• 2024-09 — Aleph Alpha exits the frontier-model race; pivots to PhariaAI + professional services + on-prem/sovereign deployment. The public Luminous API stops being marketed to new customers.
• 2024–2025 — /pricing retired (returns 404, verified 2026-06-11); pricing becomes sales-only.
• 2025-10 — Founder Jonas Andrulis steps down as CEO, becoming Chairman of the Advisory Board from January 2026.
• 2026-04-25 — Cohere agrees to acquire/merge with Aleph Alpha at a ~$20bn combined valuation, backed by ~€500M of Schwarz Group structured financing; PhariaAI folds into a “Canadian-German” sovereign-AI stack (pending approval).

The pivot, in detail

The September 2024 announcement was a rare public concession by a national-champion AI lab that the pure foundation-model business was not viable at its capital scale. Rather than continue subsidizing a premium per-token API against far better-funded rivals, Aleph Alpha reframed itself as a sovereignty and integration partner. The pricing implication is structural: a public, transparent, usage-metered rate card was replaced by a commitment-based, quote-driven enterprise model where the value metric is no longer tokens consumed but a compliant, data-resident AI deployment delivered with services. It is the inverse of the open-weight-plus-public-API playbook that Mistral runs — and a cautionary data point for any lab assuming a metered API alone can fund frontier ambition.

What’s unique : Aleph Alpha’s distinctive pricing mechanics

1. A foundation-model lab that deleted its own price card. The defining mechanic is a negative one: Aleph Alpha is the corpus’s clearest example of a lab walking away from public per-token pricing. Where peers compete on transparent $/M-token rates, Aleph Alpha retired its /pricing page (now a 404) and routes every commercial surface through a sales conversation. The pivot turned a usage-metered product into a commitment-based enterprise sale — a deliberate retreat from transparency in exchange for higher-value, stickier sovereign deployments.

2. Sovereignty as the value metric, not tokens. Post-pivot, what a buyer pays for is data residency, on-prem control, and EU-compliance — “your data never routes through a US hyperscaler.” That positioning is the price-able dimension: the same software is worth far more to a regulated German ministry than its token count would imply. Aleph Alpha monetizes regulatory posture and deployment, not raw inference — the antithesis of a pure-usage token meter.

3. Premium-specialization pricing as the historical signature. Even in its API era, Aleph Alpha never competed on price. Luminous Supreme at ~$175/M input was roughly an order of magnitude above commodity models — a bet that European-multilingual specialization justified a premium rather than a race to the bottom. That refusal to price-compete foreshadowed the pivot: when premium positioning couldn’t carry a public API against well-funded rivals, the company moved the premium into a quoted enterprise envelope instead.

Strengths & weaknesses

Billing UX : contracts over dashboards

• No self-serve billing — there is no public sign-up, no usage dashboard for new buyers, and no published rate card. Engagement begins with a “Kontakt” / sales conversation.
• Contract-based commitments — current pricing is an annual platform license plus deployment and services, negotiated per engagement rather than metered in-app.
• Legacy prepaid credits — existing Luminous API customers top up EUR credits and draw them down per token; this is the only metered surface and it is being wound down.
• Sovereign deployment controls — the “billing UX” that matters to buyers is operational: on-prem / sovereign-cloud install gives them full control over where data and compute live, rather than a spend dashboard.
• Professional-services invoicing — services are billed time-and-materials or by project milestone, typical of an enterprise/consulting motion rather than a SaaS meter.

Strategic wins : why Aleph Alpha’s pricing decisions worked

1. Quitting an unwinnable race before it bankrupted the model

Aleph Alpha’s most consequential pricing decision was to stop — to admit in September 2024 that a public per-token European LLM couldn’t out-fund OpenAI or Mistral and to redirect to a quoted sovereign-platform model. Walking away from transparent usage pricing toward commitment-based enterprise contracts preserved margin and capital that a price war would have destroyed. The candor was itself strategic: it reset expectations and refocused the business on a defensible niche.

2. Turning EU sovereignty into the value metric

By making data residency and on-prem control the thing buyers pay for, Aleph Alpha aligned price with a benefit US hyperscalers structurally cannot match for European regulated buyers. This mirrors the broader shift away from rigid per-unit licensing toward value-anchored pricing — except here the value is regulatory posture, not feature count, which is far harder for a competitor to replicate on price.

3. A multi-model platform instead of betting the company on one model

PhariaAI prices a platform that orchestrates many models rather than access to a single proprietary flagship. That decoupling means revenue no longer depends on Aleph Alpha winning a benchmark war — it depends on integration and deployment value, a steadier base. Choosing the durable value metric (deployment + sovereignty) over the volatile one (model leadership) is the discipline underneath the pivot.

Areas to improve : gaps in Aleph Alpha’s pricing approach

1. Offer even a directional public anchor

A fully sales-gated model with no published floor forces every prospect — including mid-market buyers who could self-qualify — into a sales cycle. A worked example, a “starting from” figure, or a published services day-rate would shorten evaluation and reduce the unpredictability anxiety that opaque enterprise pricing creates. Total opacity optimizes for large deals at the cost of pipeline breadth.

2. Give legacy Luminous users a clear pricing offramp

Winding down the per-token API without a published migration path or pricing bridge to PhariaAI leaves existing customers exposed. A documented transition offer — credit conversion, deployment discount, or a defined sunset date — would convert a churn risk into a controlled upsell rather than a quiet abandonment.

3. Make the sovereignty premium legible

“Contact us” hides exactly how much the data-residency and on-prem value is worth. Publishing the structure of the premium (platform vs deployment vs services), even without absolute numbers, would let buyers understand what they are paying for and why a sovereign deployment costs more than a hosted API — turning opacity into an explainable value story.

Key takeaways

1. Sometimes the smartest pricing move is to stop pricing publicly. Aleph Alpha retired its per-token rate card because the metered-API model couldn’t fund frontier ambition at its scale. Quitting an unwinnable usage-pricing race was healthier than subsidizing it.
2. Pivot the value metric, not just the price. The company didn’t discount tokens — it changed what it sells (sovereign deployment + services) so that price aligns with a benefit competitors can’t match. When the unit economics break, change the unit.
3. Sovereignty and compliance are real, price-able value metrics. For EU regulated and public-sector buyers, data residency and on-prem control are worth a premium that token counts never capture. Regulatory posture can be a pricing dimension.
4. Premium specialization rarely survives as a public API. Luminous’s ~$175/M premium positioning couldn’t carry a transparent API against funded rivals; the premium had to move into a quoted enterprise envelope to survive.
5. A retreat can still be a viable business — at a smaller scale. Around €20M ARR after a ~€500M raise is a sober outcome, but the services-and-sovereignty model is a real business, and it made Aleph Alpha an attractive merger target for Cohere’s $20bn sovereign-AI play.

UBP implications

1. Usage-based pricing needs unit economics that the metered product can actually fund. Aleph Alpha shows the failure mode: a premium per-token API that couldn’t out-spend better-capitalized rivals. UBP only works when the meter sustains the cost of delivering it — otherwise a commitment model is the honest answer.
2. When the value isn’t the consumption, don’t meter the consumption. For sovereign deployments the value is residency and control, not tokens — so a per-token meter would have mispriced the benefit. UBP designers should meter the dimension the buyer actually values, even if that means abandoning the obvious unit.
3. A transparent-to-opaque shift is sometimes a deliberate up-market move. Aleph Alpha traded public per-token transparency for higher-value, stickier quoted contracts. The trade-off — pipeline breadth for deal value — is a real lever, and the inverse of the open, transparent token pricing most AI peers chase. Know which side of it your business needs.

Sources

• Aleph Alpha official website (accessed 2026-06-11)
• Aleph Alpha /pricing — returns 404 (accessed 2026-06-11)
• TechCrunch — “German LLM maker Aleph Alpha pivots to AI support” (2024-09-05) (accessed 2026-06-11)
• The Decoder — “Aleph Alpha quits AI model race” (accessed 2026-06-11)
• TechCrunch — “Why Cohere is merging with Aleph Alpha” (2026-04-25) (accessed 2026-06-11)
• CloudPrice — Luminous Supreme rates (accessed 2026-06-11)
• CloudPrice — Luminous Base rates (accessed 2026-06-11)
• Browse the pricing blueprint corpus

Bottom line

Aleph Alpha is the corpus’s clearest study of a frontier lab abandoning public usage pricing. After raising ~€500M, it conceded in 2024 that a metered European LLM API couldn’t fund the race, retired its price card (the /pricing page is now a 404), and repositioned around PhariaAI — a sovereign-AI platform sold by quote alongside on-prem deployments and professional services. The legacy Luminous API, premium-priced at up to ~$175/M input tokens, is wound down. Pricing is now sales-only, with sovereignty and data residency — not tokens — as the value metric. In April 2026 Cohere agreed to fold it into a ~$20bn sovereign-AI stack.

Want to compare Aleph Alpha against other foundation-model labs? See Mistral AI and OpenAI, or browse the full pricing blueprint.