AnrakConductor: Deep

anrak-conductor-deepanrak/conductor-deep

Indian law · groundedProfessional+

Deep runs an AB-MCTS search over the jurist pool: it generates, scores, and refines along the most promising branches instead of drafting once. What makes it the strongest option for hard Indian-law questions is the scorer — every explored branch is checked symbolically against the retrieved authorities (citation manifest, then verification cache, then the live corpus), so the search optimises against real citation checks rather than a model's opinion of its own work. Generation-heavy and buffered; use it for the questions worth waiting on.

Modalities

text → text

Price

₹181.49 / request

flat, per call

Context

Depth

Deep search (grounded scorer)

Console

The Panel

Four independent jurists draft the same question in parallel. They are deliberately different model families, so their errors are largely independent — the condition under which fan-out plus aggregation actually beats one strong model.

Jurist	Lens	What it contributes
Athena	All-round legal drafter	Decomposes the question and covers every sub-issue methodically. Usually the most complete draft.
Vikram	Frontier reasoner	Pushes maximum depth on the hardest part of the problem. The rollout engine for deep search.
Themis	Deep reasoning specialist	Careful step-by-step chains that flag their own uncertainty rather than leaping to a conclusion.
Chanakya	Indian-law native	Trained closest to Indian legal language and drafting convention; on the pure family, the practical-implications lens.

Why no provider table. The underlying models are never disclosed on any client-bound surface — not in the API response, not in the streamed reasoning, not here. You are buying the panel and the verification around it, and we re-point individual seats as models improve without changing this id's contract or price.

Pricing

₹181.49per request

One flat charge per call, whatever the answer's length. A request to this id runs ~16 underlying model calls plus live legal research — that fan-out, not your token count, is what it costs us, so that is what we price on.

Estimate a month

Requests per month to anrak-conductor-deep.

/ month

Per request: ₹181.49
Per month: ₹54,44,700
Per 1,000 calls: ₹1,81,490

Comparing against a per-token model?

On a reference request (2,000 tokens in, 2,000 out) this price works out to an effective $206.24 / $824.95 per 1M tokens.

That number looks high next to a raw frontier model, and it should: a raw model bills you for one call, while this id bills for 16 plus retrieval and a verification pass. We publish the comparison rather than hide it — but we bill per request, so a long answer does not cost you more.

Verification

Three-cop verifier panel

A Citation Cop, a Logic Cop and a Statute Cop cross-examine the consolidated answer against the authorities that were actually retrieved. Flagged fabrications are corrected before the answer returns; if the panel cannot reach quorum, the answer is withheld rather than shipped unverified.

The verdict rides on every response as anrak.verify, the drafts and their grounded scores as anrak.orchestration, and the ordered trace as anrak.reasoning. When streaming, the same narration arrives live as delta.reasoning_content.

Quickstart

Two changes from an ordinary OpenAI call: the base_url and your ANRAK- key.

from openai import OpenAI

client = OpenAI(
    base_url="https://anrak.legal/api/v1",
    api_key="ANRAK-xxxxx",
)

resp = client.chat.completions.create(
    model="anrak-conductor-deep",
    messages=[{"role": "user", "content": "Can a Section 138 NI Act complaint be quashed if the debt is time-barred?"}],
)
print(resp.choices[0].message.content)

# What it cost, and whether a downgrade happened:
extra = resp.model_extra or {}
print(extra["anrak"]["billing"])      # {'model': 'anrak-conductor-deep', 'price': '₹181.49', ...}
print(extra["anrak"]["downgraded"])   # None unless the pipeline was downgraded

Limits and trade-offs

Professional plan or above. On a lower plan, or when the token budget is short, it silently runs as the router — read anrak.downgraded on the response to detect this.
Up to 12 buffered frontier calls per request. Latency is measured in tens of seconds, and it is by far the most expensive id in the catalog.

Best for

• The hardest questions — conflicting authority, unsettled positions, multi-issue memos
• Async and batch work where a slow, thorough answer beats a fast one
• Cases where you want the search itself grounded, not just the final check

FAQ

What exactly am I charged for?

One flat ₹181.49 per successful request, debited from your pay-as-you-go balance. Failed requests are refunded automatically — including answers withheld by the verifier, which you are never billed for even though the underlying compute was spent.

Which permissions does a key need?

model:invoke and search. A key can also be restricted to specific model ids, so a key issued for this model cannot be used to call a more expensive one.

What happens on a plan below Professional?

The request still succeeds, but it runs the standard pipeline instead of deep search, and it is billed at the lower price. The response tells you: anrak.downgraded names what you asked for, what ran, and why.

Do you train on my requests?

No. Prompts and completions on the Model API are not used to train models. The legal family additionally routes only to zero-retention providers.

Is the answer legal advice?

No. Every authority is retrieved and citation-checked, and the verifier withholds answers it cannot stand behind, but the output is a research product — a qualified professional must review it before it is relied on.