AnrakConductor
anrak-conductoranrak/conductorAnrakConductor is the default Indian legal reasoning model. A planner reads the question, live research pulls case law, statutes, court records and web intelligence, and a router decides how much deliberation the question warrants: one grounded pass for a lookup, the best two jurists for a moderate question, the whole four-jurist bench when it is genuinely hard. A three-cop verifier cross-examines the result against the retrieved authorities before it returns.
The Panel
On a hard question this id convenes the bench below; on a simple lookup it answers solo. The four 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. |
Pricing
One flat charge per call, whatever the answer's length. A request to this id runs ~5.9 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.
- Per request
- ₹39.12
- Per month
- ₹11,73,600
- Per 1,000 calls
- ₹39,120
Comparing against a per-token model?
On a reference request (2,000 tokens in, 2,000 out) this price works out to an effective $44.45 / $177.82 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 5.9 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",
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', 'price': '₹39.12', ...}
print(extra["anrak"]["downgraded"]) # None unless the pipeline was downgradedLimits and trade-offs
- Because the router adapts, latency and depth vary between calls on the same key.
- A question the classifier reads as a simple lookup gets one pass, not the bench. Use anrak-conductor-ensemble if you need the panel every time.
Best for
- • General-purpose Indian legal Q&A where you do not want to pick a depth yourself
- • Mixed workloads — the router spends tokens only on the hard questions
- • Anything that must cite real, retrieved authority rather than model memory