Copy page

Get Started

Models

Models

Hyphen serves four models behind one base URL. Name any of them in the model field of a request.

Model	Tier	Reach for it when…
m2.5	Fast / cheap	High-volume, low-latency, low-stakes calls
m2.7	Balanced	Day-to-day agent work — the sensible default
m3	Strongest / reasoning	Hard reasoning, large refactors, deep planning
text	Plain text	Straight text completion

All four are served by the same gateway and billed against the same monthly cap on your key.

Stable aliases

If you'd rather pin a tier than a version, three aliases map onto the models above. They keep working as the underlying mapping is upgraded:

Alias	Maps to	Equivalent to
hyphen-fast	m2.5	Fast / cheap
hyphen-sonnet	m2.7	Balanced default
hyphen-opus	m3	Strongest reasoning

Use whichever style you prefer — a request naming m3 and a request naming hyphen-opus hit the same model. The API reference and code samples on this site use the real IDs (m2.5 / m2.7 / m3 / text).

List what your key can reach

Code
curl https://api.hyphen-solution.com/v1/models \
  -H "Authorization: Bearer $HYPHEN_API_KEY"

Code
{
  "object": "list",
  "data": [
    { "id": "m2.5", "object": "model", "owned_by": "hyphen" },
    { "id": "m2.7", "object": "model", "owned_by": "hyphen" },
    { "id": "m3", "object": "model", "owned_by": "hyphen" },
    { "id": "text", "object": "model", "owned_by": "hyphen" }
  ]
}

Picking a model for a coding agent

Coding agents lean on the strongest model. For Claude Code, Codex CLI, Hermes, and OpenClaw we recommend m3 as the primary model, with m2.5 for cheap background calls. Each agent page shows the exact config.

Related

• Quickstart — your first request.
• Connect your coding agent — per-tool setup.
• API Reference — every endpoint, interactively.