Phind pricing

About

Phind was an AI-powered search engine and coding assistant built for developers. Founded by Michael Royzen (a UT Austin Turing Scholar, ex-Lyft / Cloudflare / Microsoft ML researcher), the company went through Y Combinator’s S22 batch under the original name “Hello” before rebranding to Phind. It positioned itself as the developer-focused answer engine: type a technical question, get an inline-cited answer with code, diagrams and (in later versions) interactive mini-apps, rather than a list of blue links.

Phind trained its own models — most notably Phind-405B (built on Llama 3.1 405B, launched September 2024, ~92% HumanEval) and the fast Phind-70B / Phind Instant — alongside reselling frontier APIs (GPT-4o, Claude 3.5 Sonnet) on paid tiers. It peaked at meaningful developer mindshare in early 2024, generated two well-received Launch HN threads (Phind 2 at 537 points in February 2025; Phind 3 at 138 points in December 2025), and on December 3, 2025 raised roughly $10.4M (reported by Axios, with Bessemer Venture Partners and Y Combinator participating).

Six weeks later it was gone. Phind ceased search operations on Friday, January 16, 2026, with about two weeks’ notice, issued prorated refunds to Pro subscribers, and deleted user data on January 30, 2026. This page is a pricing post-mortem: the historical model is reconstructed from third-party sources because the live pricing surface no longer resolves. For context on how peers are faring, browse the pricing blueprint.

Pricing summary : how Phind’s freemium developer-search plans were priced

Phind sold a freemium subscription, not usage-based pricing: a free tier with unlimited “fast” searches, a paid Pro tier (third-party listings cluster around ~$20/mo, with some later listings showing a $10 Plus tier and a $17/mo annual Pro), and a Business tier around ~$40/user/mo for teams. The meter beneath the subscription was a daily cap on “high-quality” searches (roughly 500/day on Pro) plus model access — frontier APIs (GPT-4o, Claude 3.5 Sonnet) and Phind’s own Phind-405B were gated to paid tiers.

• Free — $0, unlimited fast searches on Phind-70B, login required since April 2024.
• Pro — ~$20/mo (third-party indicative): premium models, ~500 high-quality searches/day, codebase indexing.
• Business — ~$40/user/mo (third-party indicative): no-data-retention guarantee, SOC 2, team management.

All dollar figures here are third-party indicative (aggregator listings, May 2026), not first-party confirmed — phind.com/plans now returns a Vercel DEPLOYMENT_NOT_FOUND 404, and prices were rendered client-side from a runtime API that web archives never captured, so exact final amounts are recorded as approximate rather than guessed.

What makes this different: the load-bearing fact on this page is not a price but a status — the pricing surface has been taken down because the company shut down, a freemium developer-search engine out-competed by native search in the foundation models it wrapped.

Pricing by product

The tiers below reflect Phind’s last known plan structure before shutdown. Prices are third-party indicative (aggregator listings captured May 2026) because the live plans page is dead and prices were never server-rendered into any web archive. Listings varied — most likely because Phind re-priced and renamed tiers several times across 2023–2026 — so two plausible reconstructions are shown.

Phind (Individual plans)

Phind (Business plans)

Prices are THIRD-PARTY INDICATIVE, not first-party confirmed: phind.com/plans returns DEPLOYMENT_NOT_FOUND (2026-06-08) and the app rendered plan prices from a runtime API that web archives never captured. No price is invented; ranges reflect conflicting aggregator listings.

Sales motions across products: PLG / self-serve for Free, Plus and Pro; light self-serve / sales-assist for Business.

Hidden costs : the real cost of building on a tool that can disappear

Phind’s shutdown reframes “hidden costs” entirely. There is no current bill — the product is gone, new signups are impossible, and active Pro subscribers were issued prorated refunds with a January 30, 2026 deadline to export their data. So the meaningful hidden cost was never an overage line item; it was discontinuation risk and the switching cost of rebuilding a workflow around a tool that vanished six weeks after a funding announcement.

For the record, while it operated, Phind’s freemium subscription had unusually few surprise charges: it was a flat monthly fee with a daily high-quality-search cap rather than metered overages. The “hidden cost” was soft — hitting the daily cap pushed you to slower / cheaper models or a paywall, not an invoice. The hard cost only landed at the end: any annual prepay, any team onboarding, and any prompts/answers that lived only inside Phind became stranded when the service closed.

The table below reframes “hidden costs” for a discontinued vendor — a 5-seat Business team that had standardized on Phind Pro/Business when the shutdown was announced. There is no recurring charge; the costs are one-time and operational, not billed by Phind.

The honest takeaway: with a shut-down vendor the dangerous cost is not an overage — it is the switching cost and the time spent rebuilding a workflow around a tool that disappeared.

The post-mortem lesson for buyers is to treat single-vendor AI search/coding tools as substitutable: keep prompts portable, avoid annual prepay on early-stage wrappers, and assume native search in ChatGPT/Claude/Gemini can absorb the category. See bill shock and cost unpredictability for the recurring-charge version of this risk.

Phind’s plans are no longer purchasable, so there is no live bill to estimate. To model an equivalent freemium-vs-usage AI tool, use the Phind pricing calculator or browse other pricing calculators to compare per-seat subscription vs. per-request usage costs across vendors.

Pricing evolution : from free dev search to a $10.4M raise to shutdown in six weeks

Phind’s arc ran from a free, open developer search engine to a paywalled freemium subscription, to a model company (Phind-405B), to a “mini-app” answer engine — and then to an abrupt shutdown. The pricing story is inseparable from the competitive one: every step up-market was an attempt to out-run foundation-model providers adding native search.

Cadence

Tracked range: 2023-10–2026-01. Dates and events sourced from Hacker News Launch threads, the Axios funding report, and third-party shutdown coverage; exact prices were never archived.

Notable changes

• 2023-10 — Paywall introduced on a previously free service; HN thread (2023-10-25) flagged a “steep price just for a few uses per day.”
• 2024-04-17 — Login required to use search (HN), converting anonymous usage into accounts ahead of paid conversion.
• 2024-09 — Phind-405B (Llama-3.1-405B, ~92% HumanEval) + Phind Instant launched, anchoring the Pro tier on first-party models.
• 2025-12-03 — ~$10.4M raise reported by Axios (Bessemer Venture Partners, Y Combinator) alongside the Phind 3 mini-app launch.
• 2026-01-16 — Phind ceased search operations with ~2 weeks’ notice; prorated Pro refunds; data download deadline 2026-01-30.

The shutdown in detail

Phind announced on roughly January 12, 2026 (HN, 62 points) that Friday, January 16 would be its last day of search operations, with all user data deleted on January 30. The timing — about six weeks after a publicly reported $10.4M raise — drew commentary from press (Ed Zitron noted Phind “shut down just over a month after raising over $10m”) and a wave of HN/Reddit threads. The stated cause was competitive: as ChatGPT search, Google AI Overviews, Claude web search and Perplexity matured, a standalone developer answer engine lost the moat it had as a 2023 first-mover. Search interest had reportedly fallen sharply from its early-2024 peak. Phind did the responsible things on the way out — prorated refunds and a data-export window — but the episode became a widely-cited example of the wrapper-company problem: differentiation built on top of someone else’s model evaporates when that model adds the feature natively.

What’s unique : what Phind was, and why its model mattered

1. Developer-first answer engine, not a chatbot. Phind’s wedge was technical search: ask a programming question and get an inline-cited answer with runnable code, instead of a chat transcript or a page of links. It pre-dated and arguably foreshadowed the “search-in-the-model” pattern that ChatGPT, Claude and Gemini later adopted natively — which is exactly what eroded its moat. It belongs to the AI coding cohort even though it started as search.

2. It trained its own models to anchor the paid tier. Rather than purely reselling frontier APIs, Phind shipped Phind-405B (Llama-3.1-405B base, ~92% HumanEval) and the fast Phind Instant / Phind-70B. First-party models let it offer a credible free tier (fast model, unlimited) while reserving premium APIs (GPT-4o, Claude 3.5 Sonnet) for Pro — a clean freemium fence.

3. Rich rendered output as the differentiator. Phind 2 added visual answers and diagrams; Phind 3 went further with “every answer is a mini-app,” generating an interactive React webpage per query. The bet was that presentation, not just the underlying model, could justify a subscription — a bet the market ultimately did not reward at the price point.

4. Daily-cap freemium, not usage metering. The paid fence was a daily cap on “high-quality” searches (~500/day on Pro) plus model access, layered on a flat subscription. This kept the bill predictable for users but gave Phind no usage-based upside as power users scaled — a structural contrast with the usage-based pricing cohort.

Strengths & weaknesses

Billing UX : how Phind handled plans, refunds and its shutdown

• Plans page (now dead) — phind.com/plans returns a Vercel DEPLOYMENT_NOT_FOUND 404 as of 2026-06-08; the route that hosted the plan picker no longer resolves.
• In-app self-serve checkout (historical) — paid upgrade was self-serve from the in-app account/plans surface; the standalone /plans route had been redirecting to / across 2024–2025 archive snapshots, so the plan picker lived inside the app, not on a static marketing page.
• Daily-cap usage display — Pro exposed a daily “high-quality search” allowance (~500/day); hitting it downgraded the experience to faster/cheaper models rather than charging an overage.
• Prorated refunds at shutdown — at wind-down, active Pro subscriptions were refunded on a prorated basis, and users were given until 2026-01-30 to export chat history before deletion — a clean, if abrupt, billing exit.

Strategic wins : what Phind’s pricing and product decisions got right

1. A clean freemium fence built on first-party models

By training Phind-70B/Instant for the free tier and reserving GPT-4o, Claude 3.5 Sonnet and Phind-405B for Pro, Phind could offer “unlimited fast searches” for free without bleeding API costs, while still having a credible paid upgrade. That is a textbook freemium structure — the free tier is genuinely useful but the paid models are the conversion lever. See usage-based pricing strategy for how free-tier capacity gates conversion.

2. Predictable flat pricing over speculative metering

Phind chose a flat ~$20/mo Pro fee with a daily cap rather than per-query metering. For a search tool used dozens of times a day, predictability beat precision — users never feared a surprise bill, which lowered adoption friction. This mirrors the bet many seat-based and subscription products make, trading usage upside for trust.

3. A responsible wind-down

When it closed, Phind issued prorated refunds and gave a two-week data-export window rather than cutting access overnight. In a category where outcome-based and usage models can leave customers with stranded prepaid balances, returning money pro-rata protected what reputation it could on the way out.

Areas to improve : the pricing and strategy gaps that hastened the wind-down

1. The model had no defense against native search

Phind’s core gap was strategic, not tactical: it monetized a capability — developer-grade web search — that the foundation-model providers it depended on were certain to ship natively. The fix would have been to price and position around something models couldn’t easily absorb (proprietary codebase indexing, team knowledge, persistent agents) far earlier. See choosing the right usage metric on anchoring price to a defensible value unit.

2. Flat pricing left power users’ value on the table

A flat ~$20/mo with a daily cap meant the heaviest, most-locked-in developers paid the same as light users. A usage or seat-plus-usage hybrid could have grown revenue with engagement and funded a longer runway. The bill-shock trade-off is real, but Phind erred fully toward predictability and captured none of the expansion.

3. No graceful sunset for dependent workflows

Shutting down ~six weeks after a raise, with two weeks’ notice, gave teams that had wired Phind into their workflow almost no migration runway. A longer read-only sunset, or an open-weights release of the Phind models, would have softened the blow and preserved goodwill.

Key takeaways

1. A funding announcement is not a survival signal. Phind raised ~$10.4M and shut down roughly six weeks later — buyers should weight product durability and competitive moat over recent raises when betting a workflow on a vendor.
2. Wrapper moats evaporate when the base model ships the feature. Pricing a thin layer over OpenAI/Anthropic/Google is precarious; the durable value unit must be something the foundation model can’t trivially absorb.
3. Flat freemium is great for trust, weak for expansion. Phind’s predictable ~$20/mo Pro lowered friction but captured zero upside from power users — fine for adoption, fragile for revenue under cost pressure.
4. Train-your-own-model can fence freemium cleanly. Phind-70B/Instant let Phind give away an unlimited fast tier without API bleed — a reusable pattern for any freemium AI product worried about free-tier COGS.
5. Exit conduct is part of pricing. Prorated refunds and a data-export window are the responsible counterpart to taking prepayments; vendors that take annual money owe customers a clean unwind.

UBP implications

1. Predictability-first pricing forfeits usage upside. Phind’s flat daily-cap model shows the cost of avoiding metering entirely: no revenue expansion as engagement grows, which matters most exactly when margins are under pressure.
2. Defensibility should drive the value metric. For UBP to be durable, the metered unit must map to value the vendor uniquely creates — search queries against a commoditizing model are not that; proprietary indexing or outcomes are closer.
3. Sunset terms belong in the pricing contract. As more AI vendors take prepaid/annual commitments, prorated-refund and data-export guarantees become a pricing feature buyers should demand — Phind’s clean unwind is the model to copy.

Sources

• Phind official website (accessed 2026-06-09)
• Phind pricing/plans page (now DEPLOYMENT_NOT_FOUND) (accessed 2026-06-08)
• Phind blog (accessed 2026-06-09)
• Phind on Y Combinator (accessed 2026-06-09)

Bottom line

Phind was a first-mover developer answer engine that priced cleanly — a free fast tier, a ~$20/mo Pro, a ~$40/user/mo Business tier — and still couldn’t survive the foundation-model providers adding native search. It shut down on January 16, 2026, roughly six weeks after a $10.4M raise, refunding Pro subscribers on the way out. The lesson isn’t about a price point; it’s that a pricing model is only as durable as the moat under it.

Want to compare Phind’s post-mortem against companies still standing? 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.

Pricing page returns DEPLOYMENT_NOT_FOUND

Jun 2026

Live capture of phind.com/plans returns a Vercel 404 DEPLOYMENT_NOT_FOUND (the deployment is gone, not a bot wall); WebFetch returns 403. No Wayback snapshot carries server-rendered prices (the app renders plans client-side from a runtime API), so exact final prices are recorded as unknown / third-party indicative rather than guessed.

Shutdown — search operations end

Jan 2026

Phind ceased search operations on Friday January 16, 2026, with ~two weeks' notice. Active Pro subscribers received prorated refunds; users had until January 30, 2026 to download their data before deletion. The company cited competitive erosion as native search arrived in ChatGPT, Gemini, Claude and Perplexity.

$10.4M raise + Phind 3 (mini-apps)

Dec 2025

Phind raised ~$10.4M (reported by Axios; Bessemer Venture Partners and YC participating) and launched Phind 3 — 'every answer is a mini-app,' generating interactive React mini-apps per query (Launch HN, 138 points). Six weeks before shutdown.

Phind 2 — visual answers

Feb 2025

Phind 2 shipped AI search with visual answers, diagrams and multi-step reasoning (Launch HN, 537 points). The product leaned into rich rendered output as its differentiator versus plain-text answer engines.

Phind-405B flagship model

Sep 2024

Phind launched its own Phind-405B model (built on Llama 3.1 405B, ~92% HumanEval, matching Claude 3.5 Sonnet) plus the fast 'Phind Instant' model — positioning the Pro tier around first-party models rather than reselling only frontier APIs.

Login now required

Apr 2024

Phind began requiring a login to use the previously open search (HN, 2024-04-17), tightening the free-tier funnel toward paid conversion.

Free service adds a paywall

Oct 2023

Phind, originally a free developer search engine (formerly 'Hello'), introduced paid limits — community reports (HN, 2023-10-25) note a 'steep price just for a few uses per day,' marking its shift from free tool to freemium subscription. Prices third-party indicative.