micro1 pricing

About

micro1 is a human-data and AI-training company that recruits, vets, and orchestrates domain experts to produce high-quality datasets, reinforcement-learning environments, and agent evaluations for frontier AI. Its pitch is that “a model is only as good as its data” — and that beneath every AI breakthrough sits an orchestra of human expertise. micro1 positions itself as the conductor: sourcing the sharpest minds and forging their judgment into training data that shapes how AI reasons, adapts, and integrates.

The product surface spans two pillars. The Data Engine is the human-data layer — a data platform where experts create, review, and deliver complex datasets across healthcare, legal, finance, and more, plus Merit (a data-pipeline performance dashboard) and Zara (an AI recruiter agent that sources and vets experts at high velocity). The Intelligence Platform packages that expertise into Realm (RL environments that mirror real-world scenarios), Cortex (expert-grounded agent evaluation, also sold to AI startups as a stand-alone), and a robotics-data lab for training humanoids. A separate Government track builds, deploys, and fine-tunes mission-ready agentic AI for U.S. agencies.

micro1 is a private company headquartered at 3000 El Camino Real, Palo Alto, CA, founded around 2022 and led by CEO Ali Ansari. It serves frontier AI labs, AI-native startups, and government agencies on the demand side, and a large expert community on the supply side — its government page cites 130,000+ vetted candidates across 100+ domains and 60+ languages, and claims to have created 2,000+ U.S. jobs in two months. Competitively it sits alongside human-data and RLHF vendors such as Scale AI, Surge AI, Mercor, and Turing.

The company scaled fast on the back of the post-Scale-AI realignment, in which several frontier labs sought alternative human-data vendors. Third-party reporting (TechCrunch, Sacra) puts micro1’s revenue at roughly $50M ARR by September 2025, up from about $7M at the start of that year. In September 2025 it raised a $35M Series A at a $500M valuation led by 01 Advisors — the venture firm founded by ex-Twitter executives Adam Bain and Dick Costolo — with DoNotPay founder Joshua Browder also on the board. Notably, micro1 did not always hide its prices: as recently as 2023 it operated as a transparent developer-staffing marketplace with per-engineer hourly rates printed on the site. The shift to a fully sales-led, no-public-price posture tracks its move upmarket into bespoke, enterprise-and-government human-data contracts.

Pricing summary : sales-led, custom-quoted human-data pricing with no public rate

micro1 publishes no pricing of any kind. As of 2026-06-08 the /pricing path 301-redirects to the homepage, and every buyer-facing call-to-action across all seven public surfaces is “Get in touch” or “Get access” — there is not a single dollar amount, tier, or unit rate anywhere on the site.

The model is therefore fully sales-led and custom-quoted: scope, volume, and rates are negotiated per engagement. The one quantitative billing hint comes from the government page, which says micro1’s data platform tracks “velocity, error rates, cost per task, and quality in real time” — implying engagements are measured (and likely priced) on a per-task / per-dataset volume basis rather than a flat subscription. There is no free tier and no self-serve checkout for buyers.

What makes this different: in a category where peers like Scale AI and Surge AI also quote privately, micro1 goes further by removing the pricing page entirely and routing 100% of demand through a sales conversation — a deliberate opacity that fits its enterprise-and-government, frontier-lab buyer base. This is a textbook sales-led pricing motion, but with an implied pure-usage value metric (cost per task) lurking beneath the quote. What makes it notable is the reversal: micro1 once published per-engineer hourly rates openly, then walked that transparency back as it moved upmarket.

Pricing by product

micro1 discloses no prices for any of its products. Every product line below routes to a “Get in touch” or “Get access” sales conversation; the “Price” column reflects what is publicly visible, not an estimate.

Data Engine (buyer-facing)

Intelligence Platform (buyer-facing)

Government (buyer-facing)

Sales motions across products: there are no PLG / self-serve tiers — every micro1 product is sales-led and custom-quoted. The only self-serve flow on the site is the supply-side “Apply now” path for experts joining the talent pool, which is not a buyer-pricing surface.

A note on the expert (supply) side

micro1’s experts surface advertises “competitive compensation” and “biweekly payments” at an hourly rate for the domain experts who supply the labor — but it discloses no specific hourly figure, and this is a labor-supply rate, not a product price charged to buyers.

Hidden costs : an undisclosed contract where the whole bill is the hidden cost

With micro1, the “hidden cost” problem is total: because there is no published unit rate, per-task cost, engagement minimum, or list price anywhere on the site, a buyer cannot construct even a rough bill without entering a sales conversation. There is nothing to model — the entire price is opaque by design.

What we can say comes from two anchors. First, micro1’s own government page states its platform tracks “velocity, error rates, cost per task, and quality in real time,” which is the strongest signal that engagements are metered and billed on volume of completed expert work rather than seats or a flat platform fee. A buyer’s effective bill therefore scales with task volume, task complexity (a red-teaming or chain-of-thought reasoning dataset costs more per item than simple labeling), and the QA depth micro1 layers on top. Second, the company’s now-removed 2023 marketplace billing — engineer hourly rate × hours via a Stripe subscription, “no extra fees” — shows the firm’s historical instinct was a clean, pass-through unit. Whether today’s human-data contracts preserve that simplicity or layer in management margin, minimums, and project-scoping fees is exactly what the sales gate conceals.

For buyers, the practical hidden costs to probe in a quote are: per-task rates by modality, minimum engagement size, QA/rework charges, ramp time for expert recruiting, and whether RL-environment (Realm) and agent-eval (Cortex) work is priced separately from raw data production. We deliberately omit a bill-construction table here because any dollar figure would be fabricated — no public rate exists. For the general mechanics of metering on completed work, see our guide to choosing the right usage metric and our explainer on why per-task and per-outcome billing is reshaping AI services.

Want to estimate your own micro1 bill? A dedicated micro1 pricing calculator is not yet published — and cannot be, until a public per-task rate exists to model. In the meantime, our usage-invoicing and billing-cycles guide walks through how a per-task meter is structured, and the pricing blueprint lists comparable human-data and data-platform vendors for cross-comparison.

Pricing evolution : from public hourly rates to a deliberately invisible price

micro1’s pricing history is the story of a company walking away from transparency as it moved upmarket. It began as a developer-staffing marketplace that printed prices on the page, then progressively removed them as it pivoted into enterprise human-data work for AI labs.

Cadence

Tracked range: 2023-01–2026-06. Quarters not listed were verified stable (no public price changes, no SKU additions). Note that from 2023 Q4 onward “price changes” track the disappearance of public pricing, not changes to a published rate.

Notable changes

• 2023-01 — Public per-engineer hourly rates ($28–$42/hr) and a “Pricing” nav link are live; billing is a Stripe subscription with “no extra fees.”
• 2023-10 — The “Pricing” nav link and per-profile price cards are gone; gpt-vetting (GPT-4 + Whisper) becomes the headline feature — the first retreat from transparent pricing.
• 2024-04 — The AI interviewer is publicly framed as “the first ever AI interviewer” (Hacker News, 4 points — minimal community engagement).
• 2025-09-12 — $35M Series A at a $500M valuation led by 01 Advisors; revenue reported at ~$50M ARR (up from ~$7M at the start of 2025); decisive pivot to enterprise human-data with no public price.
• 2026-06-07 — All seven public surfaces confirmed sales-gated with zero disclosed prices; /pricing redirects to the homepage.

The transparency reversal in detail

Most companies in the corpus move toward clearer pricing as they mature; micro1 went the other way. In 2023, when it sold pre-vetted engineers, the model was unusually legible for a staffing firm: you saw a developer’s hourly rate on their card, paid that rate × hours via a Stripe subscription, and were told there were “no extra fees at all.” By late 2023 the “Pricing” nav had vanished and the narrative shifted to its AI vetting engine; by the 2025 Series A the company had fully reinvented itself as a human-data engine for frontier labs — a buyer category that negotiates bespoke, NDA-bound contracts and does not expect a list price. The reversal is rational (enterprise data contracts genuinely are custom), but it means the one quantitative anchor a curious buyer can still find is a four-year-old hourly-rate screenshot in the Wayback Machine, not anything on the live site.

What’s unique : a sales-only human-data model metered on cost per task

Pricing is removed, not just gated. Plenty of enterprise vendors quote privately, but micro1 deletes the pricing surface entirely — the /pricing URL 301-redirects to the homepage, and there is no “starting at,” no tier ladder, and no published unit anywhere across seven public pages. The funnel is binary: marketing copy, then “Get in touch.” For a frontier-lab and government buyer base that expects bespoke contracts, this is a feature, not an oversight.

The implied value metric is cost per task, not seats. The only quantitative billing language micro1 publishes — on its government page — is that the data platform tracks “velocity, error rates, cost per task, and quality in real time.” That places micro1 closer to a pure-usage value metric than to the seat-based or retainer pricing common in staffing. Buyers effectively pay for completed, QA’d expert work, with complexity (red-teaming, chain-of-thought reasoning, AVLM) driving the per-task rate.

Zara, the AI recruiter, is the supply-side engine. micro1’s moat is its ability to source and vet experts at machine speed: Zara (originally “gpt-vetting,” built on GPT-4 and Whisper) interviews candidates at scale, feeding a pool of 130,000+ vetted experts across 100+ domains. The same AI-vetting capability that started the company as a developer marketplace now lets it spin up domain-specific data teams faster than human-recruiter-bound rivals.

Two-sided opacity with a one-sided exception. Buyers face a fully sales-gated experience, but the supply side is self-serve: experts can “Apply now,” take the AI interview, and onboard to biweekly-paid hourly work without ever talking to sales. micro1 advertises “competitive compensation” to that side without a published figure either — so the only transparent number the company ever shows is, ironically, neither a buyer price nor a worker wage.

A multi-product spread under one quote. The Data Engine (data production + Merit dashboard), the Intelligence Platform (Realm RL environments, Cortex agent eval, Robotics), and the Government track are all sold through the same “Get in touch” door. There is no per-product price discovery — a buyer cannot tell whether Realm and Cortex are line items or bundled until they are in a contract conversation.

Strengths & weaknesses

Billing UX : sales-gated controls — no self-serve buyer billing surface

micro1 exposes no public buyer billing UI (no checkout, no plan picker, no invoice/usage portal). The only billing-adjacent controls visible on the public site are:

• “Get in touch” / “Get access” sales gate — the single, universal entry point for buyers; every product CTA across the homepage, Data Engine, Intelligence, Cortex-for-startups, AI-Labs, and Government pages routes here. There is no online sign-up or payment.
• Merit performance dashboard — described as a data-pipeline dashboard that quantifies expert data quality, velocity, and reliability; the consumption signal a buyer would meter an engagement against, though it is not shown publicly.
• Real-time per-task tracking — the Government page states the platform tracks “velocity, error rates, cost per task, and quality in real time,” indicating engagement-level cost/usage reporting exists behind the sales gate.
• Expert talent dashboard — a supply-side control where experts check application status, manage availability, and receive biweekly payments; it governs labor payouts, not buyer billing.

Strategic wins : the decisions that compounded into $50M ARR

1. Matching the buyer’s expectation: sales-only for sales-only buyers

micro1’s frontier-lab and government buyers negotiate bespoke, NDA-bound human-data contracts — they don’t shop a pricing page. By removing pricing entirely and routing 100% of demand through “Get in touch,” micro1 maximizes deal-by-deal price discrimination and avoids anchoring large contracts to a published rate. This is the right call for the segment, even though it would be wrong for a self-serve product. For the trade-offs of going fully sales-led, see our usage-based pricing fundamentals guide and our analysis of the value-metric problem in AI pricing.

2. Turning the AI-vetting engine into a supply moat

The same gpt-vetting tool (now Zara) that launched micro1 as a developer marketplace became its durable advantage: it can interview and certify experts at machine speed, building a 130,000+ pool across 100+ domains and 60+ languages. In a market where the bottleneck is qualified human supply, owning a fast, automated vetting pipeline lets micro1 stand up domain-specific data teams faster than rivals dependent on human recruiters — a structural edge that scaled revenue ~7× in 2025.

3. Riding the post-Scale-AI realignment

When several frontier labs distanced themselves from Scale AI after Meta’s investment, the market opened for a credible alternative human-data vendor. micro1 leaned in hard, repositioning from staffing to “the human-data engine for frontier AI” exactly as demand for vendor diversification spiked — and converted that timing into a $35M Series A at a $500M valuation.

4. Cost-per-task as the value metric

By framing engagements around “cost per task” rather than seats or retainers, micro1 ties price to delivered, QA’d output — the metric labs actually care about. This outcome-aligned metering is far easier to defend in a negotiation than a headcount-based quote, and it scales naturally with a lab’s training-data appetite.

5. The government wedge

The “America’s AI workforce” / “2,000+ U.S. jobs in two months” framing opens a less price-sensitive, mission-driven buyer segment that values vetted, compliant, U.S.-based expertise. It also differentiates micro1 from offshore-heavy data vendors and creates a defensible niche where procurement is about trust and compliance, not lowest per-task price.

Areas to improve : where opacity costs more than it earns

1. Publish at least one anchor — an indicative range or engagement minimum

Total pricing silence filters out not just tire-kickers but also fast-moving startups doing real evaluation. micro1 could keep contracts custom while publishing a single anchor — an indicative cost-per-task range by modality, or a “typical engagements start at $X” floor — to pre-qualify demand and reduce sales load. Plenty of enterprise vendors disclose a starting figure without compromising negotiation leverage; the trade-offs are covered in our note on usage invoicing and billing cycles.

2. Offer a self-serve evaluation path for Cortex

Cortex-for-AI-Startups targets a segment (early agent builders) that often isn’t ready for a sales conversation. A paid, self-serve Cortex agent-eval trial — even a fixed-price starter pack — would capture that demand at the top of funnel and create a natural upgrade path into a full data-engine contract, instead of forcing a “Get in touch” on day one.

3. Surface the value metric publicly, even without a rate

micro1 already meters “cost per task” internally; explaining how an engagement is scoped and billed (the unit, the QA layers, the modalities) — without naming a price — would help buyers self-qualify and reduce friction in early sales calls. Transparency about the structure of pricing is cheaper than transparency about the number, and it would soften the perception of a value-metric black box.

4. Address the transparency-reversal narrative

Going from public hourly rates (2023) to nothing reads, to a careful observer, as a trust regression. A short public explanation — “enterprise data work is bespoke, here’s how we scope it” — would pre-empt the skepticism that the Wayback trail invites and turn the reversal into a deliberate, defensible positioning story.

Key takeaways

1. Match price transparency to how your segment actually buys. A no-pricing-page strategy is viable — even optimal — when your buyers (frontier labs, government) negotiate bespoke, NDA-bound contracts and never expect a list price. The same opacity would be fatal for a self-serve product.
2. Pricing transparency can legitimately run backward as you move upmarket. micro1 went from public per-engineer hourly rates (2023) to nothing, because its customer changed. Your pricing surface should track your buyer, not your founding ideology.
3. Pick a value metric your buyer already cares about. “Cost per task” ties price to delivered, QA’d output — the exact thing AI labs optimize — which is far easier to defend in negotiation than seats or retainers.
4. A fast supply-acquisition engine can be the real moat in a services business. micro1’s AI recruiter (Zara) let it scale qualified human supply at machine speed, which is what actually unlocked ~7× revenue growth — not the pricing model.
5. Opacity has a hidden cost: lost self-qualifying demand. Removing every anchor filters out fast-moving evaluators alongside tire-kickers; even one indicative figure or a self-serve eval tier recovers top-of-funnel without surrendering negotiation leverage.

UBP implications

1. Sales-gated does not mean usage-free. Even with no public rate, micro1’s per-task cost and quality tracking implies a usage-metered engagement under the hood — opacity at the front door, metering at the back. The pricing model and the pricing disclosure are independent choices.
2. Cost-per-task is becoming the default unit for AI human-data and agent work. As AI services shift from staffing to outcome delivery, the natural value metric is completed, QA’d work — a usage unit that scales with a lab’s training appetite, not its headcount. micro1, Mercor, and Surge are all converging on volume-of-work metering.
3. Usage-based metering can hide behind enterprise sales without losing its logic. A custom-quoted contract built on a per-task meter still aligns price with value; the enterprise wrapper just removes the public price, not the usage mechanic. For UBP strategists, the lesson is that metering and transparency are orthogonal levers you can set independently.

Sources

• micro1 homepage (accessed 2026-06-07)
• micro1 Data Engine (accessed 2026-06-07)
• micro1 Intelligence Platform (accessed 2026-06-07)
• micro1 Cortex for AI Startups (accessed 2026-06-07)
• micro1 Data Engine for AI Labs (accessed 2026-06-07)
• micro1 Government (accessed 2026-06-07)
• micro1 Expert opportunities (accessed 2026-06-07)

Browse the pricing blueprint to compare micro1 against other data-platform and AI-training pricing models.

Bottom line

micro1 is a human-data and AI-training company with a deliberately invisible price: the /pricing URL redirects to the homepage and every buyer-facing CTA is “Get in touch” or “Get access.” Pricing is fully sales-led and custom-quoted, with the only billing signal — per-task cost tracking — buried behind the sales gate. The Facts above are verified against the 2026-06-07 capture; the analysis sections await a full-research pass.

Want to compare micro1 against other data-platform and AI-training pricing? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Fully sales-gated across all surfaces

Jun 2026

Captured all seven public surfaces (homepage, data-engine, intelligence, Cortex for startups, data-engine for AI labs, experts, government). No buyer-facing price disclosed anywhere; the /pricing path 301-redirects to the homepage and every CTA is 'Get in touch' or 'Get access'. The only billing signal is a 'cost per task' metric on the government page.

$35M Series A; pivot to human-data engine for AI labs

Sep 2025

micro1 raised a $35M Series A at a $500M valuation (led by 01 Advisors), with third-party reporting (TechCrunch, Sacra) citing ARR growth from ~$7M at the start of 2025 to ~$50M by September. Positioning shifted decisively from developer staffing to a human-data engine, RL environments (Realm), and agent evaluation (Cortex) for frontier AI labs. No public price accompanied the relaunch.

GPT-vetting AI interviewer (Zara) goes public

Apr 2024

micro1 publicly marketed its AI interviewer (later branded Zara) as 'the first ever AI interviewer' on the /gpt-vetting page, surfaced on Hacker News on 2024-04-15 (4 points). The AI-recruiter engine became the company's defining capability and the wedge into AI-lab data work.

Pricing nav removed; gpt-vetting foregrounded

Oct 2023

Between the 2023-04 and 2023-10 snapshots the public 'Pricing' nav link disappeared (replaced by 'About us') and per-profile price cards were dropped. The homepage instead foregrounded 'gpt-vetting powered by gpt-4 and whisper' — the first move away from transparent, self-serve pricing toward an AI-vetting tool narrative. Source: Wayback snapshot of micro1.ai, 2023-10.

Transparent developer-staffing marketplace

Jan 2023

micro1 was a pre-vetted remote-developer marketplace. Per-engineer hourly rates were published directly on profile cards ($28–$42/hr, averaging ~$34), a 'Pricing' nav link was public, and billing ran as a Stripe subscription (hourly rate × hours, 'no extra fees'). A 1-week-free trial and a 3-month minimum applied. Source: Wayback snapshot of micro1.ai, 2023-01.