orchid-2 ยท deep

Thought for 6 s

Three clauses shift liability onto you:

Liability cap removed for data breaches

ยง9.2

Auto-renewal with 90-day notice

ยง14.1

Indemnity extends to subcontractors

ยง11.4

MSA_v3.pdf

Order_Form.pdf

+12

14 sources

Orchid reads every page, checks every claim and shows you where each answer came from โ€” so the work you hand off comes back ready to sign.

Trusted by teams who read the fine print

  • lumora

  • Fieldnote

  • KESTREL

  • Meridia

  • tessellate

  • Halcyon

  • Copperleaf

  • VERITY

  • lumora

  • Fieldnote

  • KESTREL

  • Meridia

  • tessellate

  • Halcyon

  • Copperleaf

  • VERITY

Platform

Start in Studio.
Ship it with the API.

Studio and the API run the same model with the same settings โ€” so the prompt you refine on Monday is the one your product calls on Thursday.

Studio and the API run the same model with the same settings โ€” so the prompt you refine on Monday is the one your product calls on Thursday.

Untitled prompt

orchid-2 ยท deep

Which planets in our solar system have rings? Answer as a short table.

Thought for 4 seconds

Planet

Rings

Jupiter

Faint, dusty

Saturn

Bright, icy

Uranus

Narrow, dark

Neptune

Clumpy arcs

Orchid Studio

Test prompts, compare answers side by side and prototype whole workflows before you write a single line of code.

import orchid


client = orchid.Client()

answer = client.reason(

    model=orchid.models.DEEP,

    input=contract,

    cite_sources=True,

)

200 OK

412 ms

7 citations

Orchid API

Call the model over a clean REST API with full control of reasoning depth, tools and output format. Python and TypeScript SDKs included.

same model

Documents

Bring the whole archive.
Leave with the answer.

Contracts, filings, transcripts and research โ€” Orchid reads them end to end, connects what one document says to what another implies, and shows exactly where every finding came from.

Contracts

Research papers

Financial filings

Internal memos

Product specs

Call transcripts

Q4_board_deck.pdf

Vendor_MSA_v3.docx

ledger_q3.csv

Connections found

across 14 files

conflicts with

Vendor MSA ยง9.2

Order form ยง3

matches

DPA ยง6

Data policy 4.1

not supported by

Board deck p.12

Ledger Q3

Method

It thinks before it speaks.

Every answer passes through four quiet stages. Watch each one unfold in Studio, or ask the API to return the full trace.

Reasoning trace

5.0 seconds ยท orchid-2

1.2 sec

01

Read

Takes in every page, table and footnote โ€” up to a million tokens at once.

1.2 sec

01

Read

Takes in every page, table and footnote โ€” up to a million tokens at once.

2.8 sec

02

Reason

Breaks the question apart and works through each piece before it commits.

0.6 sec

03

Verify

Checks each claim against the source and flags anything it canโ€™t support.

0.4 sec

04

Answer

Replies plainly, with citations you can click and a trace you can audit.

Capabilities

Built for depth,
not just speed.

The details that make an answer worth trusting are built into the model, not bolted on afterwards.

1,000,000

Tokens in a single prompt

A million tokens of context

Drop in a codebase, a data room or a year of research notes. Orchid keeps every page in view while it reasons.

Revenue grew 18% year over year, driven by enterprise renewals [3]

Annual report 2025 ยท p. 42

Cites its sources

Every claim links back to the passage it came from, so you can check the work in a click.

Data retention

0 days

Hosting region

US

EU

Train on my data

Private by default

Zero data retention, SOC 2 Type II and regional hosting in the US and EU. Your data never trains our models.

Agent run

3 steps ยท 2.6 s

search(q3_revenue)

120 ms

run(forecast.py)

2.1 s

fetch(sec.gov/10-K)

340 ms

Tools and agents

Call functions, browse, run code and hand work between steps โ€” with every action logged and reversible.

DE

Die Haftung des Auftragnehmers ist ausgeschlossen.

nuance kept ยท legal register

PT

A responsabilidade do contratado fica excluรญda.

Fluent in 40+ languages

Reads a contract in German and answers your question in Portuguese without losing the nuance.

Benchmarks

Measured,
not marketed.

We publish how Orchid performs on the work it is built for โ€” and where it still falls short.

Internal evaluations, October 2026. Full methodology in the model card.

94.2%

Graduate-level reasoning

Accuracy on a held-out set of expert science questions.

3.1ร—

Fewer factual errors

Compared with Orchid 1 on long-document question answering.

380 ms

Median first token

Even with deep thinking on, answers start almost at once.

380 ms

Median first token

Even with deep thinking on, answers start almost at once.

99.95%

API uptime

Measured across all regions over the last twelve months.

99.95%

API uptime

Measured across all regions over the last twelve months.

Voices

Loved by people who
read the fine print.

โ€œOrchid doesnโ€™t just answer faster. It answers like someone who actually read the whole thing โ€” and then tells you exactly where to look.โ€

Priya Natarajan

VP of Product, Lumora

โ€œWe gave Orchid eleven years of vendor contracts. It found the three clauses our outside counsel had missed โ€” and showed exactly where they were.โ€

Amara Okafor

Head of Legal Operations, Meridia

โ€œItโ€™s the first model our analysts trust without re-checking every number. The citations alone changed how we write research.โ€

Daniel Reyes

Director of Research, Halcyon Capital

Pricing

Simple plans that
grow with you.

Start free, upgrade when Orchid becomes part of how you work. Every plan includes the full model โ€” no lighter versions.

Seed

For curious individuals exploring what careful AI can do.

$0

forever

50 deep-think prompts a day

Studio with saved prompts

200K-token context

Community support

Bloom

Most popular

For professionals who need answers they can defend.

$24

per month

Unlimited deep-think prompts

Full 1M-token context

API access with $10 monthly credit

Citations and reasoning traces

Priority support

Grove

For teams and enterprises working with serious data.

Custom

billed annually

Everything in Bloom

SSO, SCIM and audit logs

Zero data retention

Regional hosting in the US or EU

A dedicated success partner

Prices in USD. Education and non-profit discounts available on request.

FAQ

Questions,
answered.

Canโ€™t find what youโ€™re looking for? Our team usually replies within a working day.

What makes Orchid different from other models?

Orchid is tuned for careful reading rather than quick replies. It works through a question before it answers, checks its claims against your sources and shows you where each one came from.

Is my data used to train your models?

Never. Prompts and files are processed only to answer you. On Grove, zero data retention means nothing is stored at all once a response is complete.

How long can my documents be?

Bloom and Grove support up to a million tokens in a single prompt โ€” roughly 2,500 pages. Seed supports 200,000 tokens.

Can I use Orchid through the API and Studio at once?

Yes. Both share the same model and settings, so a prompt you refine in Studio behaves exactly the same when you call it from your code.

Which languages and file types are supported?

Orchid reads and writes more than 40 languages, and understands PDF, Word, spreadsheets, Markdown, code and plain text โ€” including scanned documents.

Can I cancel or change plans anytime?

Of course. Upgrades take effect immediately and downgrades at the end of your billing period. There are no cancellation fees.

Start growing something
thoughtful.

Bring your hardest documents and your most careful questions. Orchid is free to try, and itโ€™s ready when you are.

No credit card required

Create a free website with Framer, the website builder loved by startups, designers and agencies.