Z.AI · GUIDE DE MODÈLE POUR HERMES

GLM-5 API avec Hermes Agent : tâches et prix

GLM-5 is a general purpose model for Hermes Agent. It is a practical starting point when a task needs reasoning, code, and tool calls but does not require a specialist visual or persistent workflow model.

Ouvrir le Playground →Ouvrir le Dashboard →

Flux Hermes Agent reliant les tâches GLM, les fichiers, les outils et une validation humaine

Informations techniques pour Hermes

General purpose agent

These facts show where GLM-5 fits inside Hermes and which details change the cost and review path.

InputText

OutputText

Best roleMain Agent or router

ToolsFunction calls and structured output

Price noteUse the current Hermes API price card

Typical jobResearch, coding, triage, and tool calls

Points de vigilance

It is a general model, not a specialist vision model.
A long task still needs a stop condition and a verification step.
Do not let it change production data without an approval gate.

À vérifier avant de lancer l’Agent

Define the output format before the first tool call.
Allow only the tools needed for this request.
Ask for sources, tests, or a confidence note.
Route specialist work instead of forcing one model to do everything.

Choisissez selon le travail, pas seulement le nom.

Use it for balanced work

Research, coding, tool calls, structured answers, and tasks that do not need a specialist model.

Do not use it for every extreme case

For visual coding, persistent OpenClaw tasks, or huge repositories, use the model built for that job.

A simple GLM-5 routing rule

Start with GLM-5 for bounded work. Route visual tasks to GLM-5V-Turbo, persistent tool chains to GLM-5-Turbo, and project scale coding to GLM-5.2 or GLM-5.3.

Tâches adaptées à Hermes

Choisir le rôle selon le travail

Les modèles GLM n’ont pas tous le même rôle. Choisissez la route après avoir décrit la tâche.

Research and synthesis

Search approved sources, compare findings, and return a clear answer with evidence.

Everyday coding

Implement a bounded change, run checks, and explain the result.

Structured operations

Classify requests, call a tool, and return JSON for the next workflow step.

Model routing

Read the task and decide whether a specialist model should handle the next step.

A user needs a clear answerGLM-5 as the main Agent

It can research, explain, and return a structured answer.

A request needs one or two toolsGLM-5 with a small tool set

Keep the tool list short so the Agent can verify each result.

The next step is visual or project-wideRoute to GLM-5V-Turbo or GLM-5.2

A specialist model will see more of the required state.

Flux Hermes

GLM-5 is a useful default, not a permanent lock in.

Hermes can begin with GLM-5, measure what the task needs, and route visual, long context, or persistent work to a specialist model.

01 · Understand

Read the request and identify the goal, data, tools, and risk.

02 · Act

Complete the bounded task with the available tools.

03 · Check

Return structured output, tests, sources, or uncertainty.

04 · Route

Move the next step to a more specialized model if needed.

Flux de recherche GLM, des fichiers et graphiques au rapport vérifié

Practical case

A small team uses GLM-5 to triage support requests

Hermes receives customer requests from several channels. GLM-5 classifies each request, extracts the product and urgency, checks the internal help tool, and returns a structured record. Requests that require a screenshot go to GLM-5V-Turbo. Requests that require a large repository change go to GLM-5.2. Hermes keeps the first model as the low friction entry point and routes only the work that needs more capability.

Ce que fait Hermes dans ce cas

01Brief

Hermes receives a support request and the customer record.

02Triage

GLM-5 extracts the product, urgency, language, and missing facts.

03Tool check

Hermes calls the approved help-center and account tools.

04Route

Visual bugs go to GLM-5V-Turbo; repository changes go to GLM-5.2.

Résultat

The team keeps GLM-5 as the low-friction first step and pays for specialist models only when the request needs them.

Vérification humaine

A person checks refunds, account changes, policy exceptions, and any reply that could create a legal or financial commitment.

Décision de coût

Use GLM-5 as the balanced starting point.

A general model is useful when the task shape is not known yet. Once Hermes sees repeated patterns, route those patterns to a cheaper or more specialized model.

Un exemple simple de coût

For a short support request, keep the prompt small and use GLM-5 once. If the same request pattern repeats, move extraction to a lower-cost model and keep GLM-5 for the uncertain cases.

Use GLM-5 in Hermes

Select glm-5, define the output format, and give Hermes a clear rule for when to route the next step to another model.

Ouvrir le Playground →Obtenir la clé API →

curl https://api.hermes-ai.net/api/v1/chat/completions \
  -H "Authorization: Bearer $HERMES_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5",
    "messages": [{"role":"user","content":"Run the approved task and return a short verification report."}]
  }'

FAQ

Questions fréquentes

What is GLM-5 best for?

It is a balanced choice for reasoning, coding, research, structured output, and tool calls.

Should GLM-5 be the default Hermes model?

It can be a useful starting point, but repeated workflows should move to the model with the best cost and capability fit.

How is GLM-5 different from GLM-5.1 and GLM-5.2?

GLM-5.1 focuses on sustained autonomous work, while GLM-5.2 focuses on project scale context and cross file engineering.

Can it call tools?

Yes. Hermes should still limit tool permissions and require a verification result.

Why use GLM-5 instead of a cheaper model?

Use it when the task needs more reasoning than a small model but does not need a specialist workflow model.

Lire la documentation officielle ↗Ouvrir le Playground ↗

HERMES API · GLM-5

Use GLM-5 in Hermes

Select glm-5, define the output format, and give Hermes a clear rule for when to route the next step to another model.

Ouvrir le Playground →Ouvrir le Dashboard →