Athina AI pricing

About

Athina AI is a collaborative AI development platform that lets product and engineering teams build, test, evaluate and monitor LLM features in one workspace. It spans the full lifecycle — prompt management and a multi-prompt playground, dataset experimentation with SQL notebooks, an evals suite (online, offline, preset and custom evaluators), agentic “Flows,” annotation tooling, and production monitoring with tracing and analytics. The product is explicitly designed for both technical and non-technical collaborators to work on the same experiments, datasets and prompts.

Athina is used by AI teams including Perplexity, Meesho, Doximity, PhysicsWallah, You.com, Vetted, Richpanel and CourtCorrect, who cite it for LLM observability, dataset curation with inter-annotator agreement, and end-to-end experimentation. The platform positions itself against the LLM observability and eval tooling category, differentiating on breadth (prototype-to-production coverage), collaboration, and data-privacy options including self-hosted VPC deployment and SOC-2 Type 2 compliance.

Athina AI is a private company in Y Combinator’s Winter 2023 (W23) batch, founded by Shiv Sakhuja (CEO) and Himanshu Bamoria. It raised a $3M seed round announced 15 November 2024, with participation from Y Combinator, Flourish Ventures, the Kleiner Perkins Scout Fund, GSF and Bharat Founders Fund, plus angels including Perplexity CTO Denis Yarats and Snorkel AI CEO Alex Ratner. Public revenue, valuation and headcount figures are not disclosed on its pricing or marketing surfaces, and — notably — the company used to publish a self-serve paid price ($99–$199/mo through mid-2024) but has since gated the entire paid ladder behind sales (see Pricing evolution).

Pricing summary : How Athina’s freemium tiers and execution-credit metering work

Athina uses a freemium model with execution-credit metering across two visible dimensions:

1. Plan tier (gated above free): A free Starter tier (10,000 logs/mo, unlimited prompts, advanced analytics) is the only tier with a public price. Pro is quote-only (“Let’s talk”) and unlocks unlimited logs, evals, datasets and team seats plus a GraphQL API. Enterprise is custom-priced and adds self-hosted deployment, SOC-2 Type 2 certification, advanced access controls and custom-model support.
2. Execution credits (usage meter): Paid usage is tracked in execution credits where 1 execution = 1 credit, irrespective of token count or where the execution runs. New accounts start with 500 free credits. Prompt runs, flow steps, offline evals, and Datasets dynamic-column/experiment cells each consume a credit; logs, online evals and annotations do not.

What makes this different: Athina decouples its headline log allowance from its compute meter — logs are free-tier capacity while executions (the actual eval/experiment compute) are the credit-metered dimension, and a single dataset operation can burn dozens of credits at once (50 rows = 50 credits). This puts it in the usage-based pricing camp on usage while keeping the seat/plan ladder sales-gated — a structure worth weighing against usage-based pricing for SaaS AI. For teams selecting a meter, choosing the right usage metric matters as much as the rate.

Pricing by product

Athina platform (plans)

Sales motions across products: PLG / self-serve for the free Starter tier; sales-led (quote-only) for Pro and Enterprise.

Execution-credit metering

Paid usage is denominated in execution credits, documented in the Athina docs. Every account starts with 500 credits on sign-up.

Hidden costs : what Athina teams actually pay beyond the free tier

The advertised free Starter tier understates real-world cost because (a) the paid Pro/Enterprise tiers are quote-only with no public number, so the moment you outgrow free you enter a negotiation rather than a checkout, and (b) batch Datasets operations multiply execution-credit burn — a single dynamic column over a large dataset can consume hundreds of credits at once.

Where the meter bites — execution-credit archetypes

What is not a hidden cost (and is worth crediting Athina for): logs — the heaviest everyday operation — are exempt from the credit meter and capped only by the 10k logs/mo free allowance. Online evals and annotations are also credit-free. The credit burn is concentrated in discrete, batchy compute (prompt/flow/eval/dataset cells), which is at least attributable per operation.

Want to estimate your own Athina bill? Use the Athina pricing calculator to model execution-credit consumption once published rates are available, or browse the pricing calculator hub for related tools.

Pricing evolution : from free observability to credit-metered execution

Athina’s pricing has moved through three distinct eras in roughly two years: a hallucination-monitoring product with sales-led tiers (early 2024), a brief experiment with a published self-serve paid price ($99 then $199/mo, mid-2024), and the current freemium-plus-execution-credits structure where $0 is the only public number (late 2024 onward).

Cadence

Tracked range: 2024-02 to 2026-06. Wayback main-surface snapshots 2024-02 through 2025-08 (16 distinct captures); live capture 2026-06-04. Quarters without pricing activity omitted.

Notable changes

• 2024-06-17 — First published self-serve price appears: Starter $99/mo (100k logs/mo, 3 seats, custom evals, synthetic data, Slack support), alongside Free $0/mo, Pro “Let’s Talk” and Enterprise Custom. Coincides with a reposition from monitoring to experimentation. (Wayback 20240617235741)
• 2024-08-03 — Self-serve Starter price roughly doubles to $199/mo (now 100k logs/mo with continuous evaluation and dedicated Slack support). (Wayback 20240803025423)
• 2024-09-10 — The $199/mo published tier is removed entirely, collapsing the ladder to Free $0/mo + Pro “Let’s Talk” + Enterprise Custom — the structure still live today. From this point, $0 is Athina’s only public price. (Wayback 20240910123620)
• 2024-11-15 — Athina announces a $3M seed (YC W23; angels from Perplexity and Snorkel AI); a funding banner and the current “collaborative AI development platform” homepage appear in the December snapshot. (CB Insights)
• 2025 (full year) — No pricing-page changes detected: the 2025-05 and 2025-08 snapshots are digest-identical, and the execution-credit meter (1 execution = 1 credit, 500 free) is documented in the developer docs rather than on the pricing page.

The disappearing paid tier, in detail

The most consequential change in Athina’s pricing history is not a price increase but a price withdrawal. For roughly one quarter — June through August 2024 — Athina ran a conventional self-serve SaaS ladder with a real number a buyer could act on without talking to sales: $99/mo, then $199/mo, for a 100k-logs Starter tier. By the September 2024 snapshot that tier was gone, and the published ladder collapsed to a free plan plus two quote-only tiers (“Let’s Talk” / “Custom”).

Two things make this notable. First, the direction is unusual: most LLM-tooling vendors add self-serve pricing as they mature, not remove it. Second, the withdrawal closely preceded the company’s $3M seed and platform rebrand — the move from a self-serve experimentation tool toward an enterprise-leaning “collaborative AI development platform” courting customers like Perplexity and Doximity. The new execution-credit meter (introduced in the docs, not the pricing page) replaced the simple seat-and-log tiering with a usage primitive better suited to bespoke enterprise contracts. No community backlash to the change was found (see community signal below) — the removal happened quietly, consistent with a deliberate shift up-market rather than a reaction to user pushback.

What’s unique : token-blind credit metering and a fully sales-gated paid ladder

1. Token-blind execution credits. Athina charges 1 credit per execution regardless of token count — a rare flat-per-operation meter in an LLM-tooling market that usually bills per token. A GPT-4o eval over a 200-token response and one over a 20,000-token response cost the same single credit, trading precision for forecastability. The cost only scales with how many discrete operations you run, not how big each one is.

2. Logs are deliberately outside the meter. The most common operation in an observability product — ingesting a production log — consumes zero credits. Online evals and annotations are exempt too. Athina reserves its meter for discrete experimentation compute (prompt runs, flow steps, offline evals, dataset cells), so the continuous-monitoring use case stays predictable and cheap. This is the opposite of per-log billing common elsewhere in LLM observability.

3. A pricing ladder that went less transparent over time. Almost uniquely among the LLM-tooling vendors in this corpus, Athina removed a published self-serve price ($199/mo) rather than adding one — withdrawing it in September 2024 as it moved up-market ahead of its $3M seed (see Pricing evolution). The result is a fully sales-gated paid ladder where the only public number is $0.

Strengths & weaknesses

Community signal. Searched HN Algolia, Reddit and major outlets for pricing-specific reaction (June 2026). No trust event meets threshold: the largest Athina HN story is a 98-point Show HN for its open-source RAG cookbooks (2024-12-04, 27 comments) — about the GitHub repo, not pricing; other Athina Show HN posts sit at 6–14 points. No Reddit thread or press coverage discusses the pricing changes themselves, and the September-2024 removal of the $199 tier drew no visible backlash. Press coverage is limited to the November 2024 seed announcement (American Bazaar, YourStory, Inc42). Net: pricing has evolved quietly, with no community controversy on record.

Billing UX : credit balance, top-ups and execution-level metering

• Credits settings page — shows the current execution-credit balance and lets users purchase additional credits directly from the in-app Settings → Credits screen.
• Execution-level metering — usage is tracked per execution (1 credit each), so consumption is attributable to specific prompt runs, flow steps, offline evals and Datasets operations rather than to opaque token totals.
• Free-credit starter grant — every new account is provisioned with 500 execution credits on sign-up, separate from the plan’s 10,000-logs-per-month allowance.
• Sales-gated upgrade path — Pro and Enterprise upgrades route through “Book a demo” / “Book a call” CTAs rather than a self-serve checkout.

Strategic wins : decisions that strengthen Athina’s pricing

1. Token-blind credit metering keeps bills predictable

By charging 1 credit per execution regardless of token count, Athina gives teams a usage meter they can forecast without modeling token volume — a contrast to the per-token billing common across usage-based pricing in AI tooling.

2. A genuinely generous free tier lowers adoption friction

The Starter plan ships 10,000 logs/mo, unlimited prompts and 500 execution credits at no cost, letting teams prove value before any sales conversation. This entry mirrors the land-and-expand motion seen across developer tooling and the value-metric thinking behind modern AI pricing.

3. Credit-free logs decouple observability from compute

By exempting logs, online evals and annotations from the credit meter, Athina keeps continuous production monitoring cheap while reserving the meter for discrete compute. That separation makes the heaviest everyday use case — logging — predictable.

4. Trading a self-serve tier for credits suited the move up-market

Removing the $199/mo Starter tier in September 2024 looks counterintuitive, but it aligned the pricing with the company’s enterprise reposition (and $3M seed) — replacing a rigid seat-and-log tier with an execution-credit primitive that flexes across bespoke contracts with customers like Perplexity and Doximity. For an enterprise-led motion, a usage meter plus “Let’s talk” is a defensible structure.

Areas to improve : gaps in Athina’s pricing transparency

1. Publish at least an entry price for Pro

The entire paid ladder is quote-only, which adds friction for self-serve teams that have outgrown the free tier. Publishing even a starting Pro price (or per-credit top-up rate) would reduce procurement drag — a recurring theme in pricing AI products with unpredictable costs.

2. Disclose the per-credit top-up rate

Athina documents that additional credits can be purchased, but no rate appears on the pricing page or docs. Surfacing a per-credit price would let teams model execution cost before committing.

3. Clarify where the free 500 credits sit relative to plan limits

The relationship between the 500 starter credits and the Starter plan’s 10,000-logs allowance is documented only in the developer docs, not the pricing page. Consolidating both dimensions on the pricing surface would reduce confusion for evaluators.

4. Reconsider whether removing self-serve pricing cost mid-market deals

Athina once let a team upgrade for $99–$199/mo without a sales call; today the smallest paid step is a quote. For mid-market teams that have outgrown the free tier but are below enterprise-procurement scale, that gap can push them toward competitors with published self-serve tiers. Re-introducing a transparent mid-tier (even credit-pack pricing) would recapture that segment without undermining the enterprise motion.

Key takeaways

1. A free tier can lead with capacity, not compute. Athina’s free plan gives away 10,000 logs/mo while metering the actual eval/experiment compute in credits — separating the cheap dimension from the costly one.
2. Token-blind metering trades precision for predictability. Charging per execution rather than per token makes bills forecastable but can over- or under-charge relative to true compute cost.
3. Pricing transparency can move backwards. Athina published a $99→$199/mo self-serve tier in mid-2024, then withdrew it by September 2024 — a reminder that “what’s the price?” can have a different answer at different points in a company’s life. Always check the current pricing surface, not last year’s.
4. Pricing changes can be invisible to the market. The removal of a published paid tier generated zero community reaction — no HN thread, Reddit post or press coverage — showing that even material pricing shifts at an early-stage vendor often happen entirely under the radar.
5. Funding events and pricing repositions cluster. Athina’s move up-market (dropping self-serve, adding execution credits) lined up with its $3M seed and rebrand — a pattern worth watching when a vendor raises: the next pricing change is often a move toward “Let’s talk.”

UBP implications

1. Per-operation credits are a viable alternative to per-token billing. Athina shows that a flat 1-credit-per-execution meter can work for LLM tooling where token counts vary wildly per call — buyers get a forecastable unit even when underlying model costs swing.
2. Decouple the free dimension from the metered dimension. By making logs free and metering only experimentation compute, Athina lets the cheap, high-volume operation drive adoption while the costly, discrete operation drives revenue. For usage-based pricing design, choosing which dimension to meter matters as much as the rate.
3. A usage meter can be the bridge from self-serve to enterprise. Replacing a fixed self-serve tier with execution credits gave Athina a primitive that flexes across both PLG free users and bespoke enterprise contracts — useful when a company is repricing as it moves up-market, but at the cost of the published mid-tier that self-serve buyers rely on.

Sources

• Athina AI pricing (homepage #pricing anchor) (accessed 2026-06-04)
• Athina docs — Credits (execution-credit billing) (accessed 2026-06-04)
• Athina docs — Self-hosting (Enterprise deployment) (accessed 2026-06-04)
• Athina docs overview (accessed 2026-06-04)

Bottom line

Athina AI pairs a genuinely generous free tier — 10,000 logs a month, unlimited prompts and 500 starter credits — with a token-blind execution-credit meter that makes paid usage unusually predictable. The catch is transparency, and it has gotten worse over time: Athina briefly published a $99→$199/mo self-serve tier in mid-2024, then removed it as it moved up-market ahead of its $3M seed. Today both Pro and Enterprise are quote-only, so the moment a team outgrows free, the price becomes a conversation rather than a number.

Want to compare Athina against other LLM-observability 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.

Current state: three-tier freemium + execution-credit metering

Jun 2026

Live capture confirms Starter Free (10k logs/mo, unlimited prompts, advanced analytics), quote-only Pro ('Let's talk'), and custom Enterprise. Paid usage metered in execution credits (1 execution = 1 credit), with 500 free credits on sign-up; logs, online evals and annotations are credit-free.

Three-tier freemium stable; execution-credit meter documented

Aug 2025

Pricing page unchanged across 2025 (digest-identical 2025-05 and 2025-08 snapshots): Starter Free (10k logs/mo) / Pro 'Let's talk' / Enterprise Custom. Paid usage is now metered in execution credits (1 execution = 1 credit, 500 free on sign-up) per the docs. Source: Wayback 20250805144542.

$3M seed funding banner; rebrand to 'Ship AI to prod 10x faster'

Dec 2024

Site adds a 'Athina AI raises $3M in new funding' banner (seed announced 2024-11-15) and rebrands to the current collaborative-platform homepage. Three-tier structure (Starter Free / Pro 'Let's talk' / Enterprise Custom) is retained; SOC-2 Type 2 now called out on Enterprise. Source: Wayback 20241202102459.

Published paid tier removed — collapse to free + two quote-only tiers

Sep 2024

The $199/mo Starter tier disappears entirely. Pricing collapses to the current shape: Free $0/mo (10k logs/mo), Pro 'Let's Talk' (unlimited logs/evals/datasets/seats, GraphQL API), Enterprise Custom (self-hosted, RBAC, custom models). From here on, $0 is the only public number. Source: Wayback 20240910123620.

Self-serve Starter price raised $99/mo → $199/mo

Aug 2024

The paid Starter tier roughly doubles to $199/mo (100k logs/mo, continuous evaluation, unlimited experiment runs, custom evals, dedicated Slack support). Free $0/mo, Pro 'Let's Talk' and Enterprise Custom unchanged. Source: Wayback 20240803025423.

First published self-serve paid price: Starter $99/mo

Jun 2024

Platform repositions from monitoring to experimentation ('Your AI team needs · Experiments'). Pricing gains a real number: Free $0/mo, Starter $99/mo (100k logs/mo, 3 seats, custom evals, synthetic data, Slack support), Pro 'Let's Talk' (1M logs), Enterprise Custom. Source: Wayback 20240617235741.

Free tier formalized with limits; three sales-led tiers above it

Apr 2024

Starter $0/mo gains explicit limits (10k logs/mo, 30-day retention, 1 seat, prompt management). Monitor (100k logs, 90-day retention, 3 seats) and Evaluate (1M logs, 100k automatic evals, GraphQL API) become 'Contact Us'; Enterprise stays Custom. Still no published paid price. Source: Wayback 20240414141110.

Hallucination-monitoring product: free tier + three 'Contact Us' tiers

Feb 2024

Earliest archived pricing. Athina is positioned as an LLM hallucination/safety monitor ('Your AI needs · Monitor your LLMs in production'). Four tiers: Starter $0/mo (up to 10k inferences, analytics), then Monitor, Evaluate and Enterprise all priced 'Custom / Contact Us'. No self-serve paid number. Source: Wayback 20240223175620.