Hugo

Four models. One voice.

Hugo is the venode language model. A cloud service shipped in purpose-trained variants, each tuned for a specific class of work. One API, one persona, the right variant for the job.

Use Hugo See how it works 

  
  hugo://session/0001
  live

~ hugo open draft-03.md
· read 412 lines · 32K context · local
· voice: matched to draft-02.md
› ready.

What we build for.

01 · principles

Purpose-trained variants.

Four variants on one base. Daily, Analytics, Engineering, and Security each carry their own fine-tune and their own evals. Pick the variant for the job.

Long context, every variant.

32K tokens by default across all four variants. Hugo loads the whole document, then the citations, the linked notes, and the repository before it answers.

One API. One persona.

Switch variants with a single parameter. The system message, the cadence, the restraint stay the same. Hugo always sounds like Hugo across every variant.

02 · the model family

HugO, four variants.

Hugo is the venode language model, served from the cloud and exposed through one API. Four variants share the base. Each one carries its own fine-tune, its own eval set, and its own system message. Pick the variant for the work in front of you.

  1. DLY Hugo Daily. Everyday writing. Notes, drafts, replies, summaries. The default variant, tuned for plain reading and editorial restraint.
  2. ANL Hugo Analytics. Computational and quantitative work. Tabular reasoning, formula derivation, statistical write-ups, dataset summaries with citations.
  3. ENG Hugo Engineering. Code and engineering. Reviews a diff, explains a stack trace, writes the test before the patch. Tuned on a corpus of real review comments.
  4. SEC Hugo Security. Threat modelling and cyber security. Reads a config, names the attack surface, drafts the response runbook. Refuses ambiguous offence work by default. 

  
  POST api.venode.io/v1/chat
  live

$ hugo --variant ENG \
    --file fire-engineering-report-03.md \
    --task review

· routed to hugo-engineering-1
· loaded 412 lines, 1,841 tokens
· latency: 312 ms first token

› [14] "lead with" → "open on"
› [27] "smoke result, then code" →
         "smoke result. Code last."
› [54] tighten run-on (12 words → 9)

· 6 edits applied · 0.28 s
› ready.

How HugO compares.

03 · honest table
Hugo General cloud model Domain copilot
Purpose-trained variants Four variants One model One vertical
Switch variants with one parameter Single API Single API Per-vendor API
Consistent persona across variants One voice Drifts on system prompt Voice per vendor
Long context, every variant 32K standard Tier-gated Cursor-local
Editorial restraint by default House style off House style on House style on

HugO is ready when you are.

Open in the browser. Free to use, no signup, while we tune it.

Use Hugo Talk to us about HugO+