Models

Dahl exposes models through the OpenAI-compatible GET /v1/models endpoint. Always read IDs from the live response — IDs can change as the Gonka network adds or rotates capacity.

curl https://inference.dahl.global/v1/models

Currently available

These models are typically returned from production (verify with GET /v1/models):

Model ID	Notes
`MiniMaxAI/MiniMax-M2.7`	Default recommendation — general chat and coding
`moonshotai/Kimi-K2.6`	Long-context and reasoning workloads
`deepseek-ai/DeepSeek-V4-Flash-0731`	Fast coding, reasoning, and agent workloads (1M context)
`zai-org/GLM-5.2`	Coming soon — not available for inference yet

The landing page shows the same model IDs in copy-paste curl examples (MiniMax, Kimi, and DeepSeek tabs). Only the model field changes between them.

Default choice

Start with MiniMax M2.7 (MiniMaxAI/MiniMax-M2.7). It is the default in the Dahl chat UI and matches the examples on inference.dahl.global.

Using a model in chat completions

Set the model field to the exact id string from GET /v1/models. All requests require Authorization: Bearer (see API keys). MiniMax M2.7:

curl https://inference.dahl.global/v1/chat/completions \
  -H "Authorization: Bearer $DAHL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMaxAI/MiniMax-M2.7",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Kimi K2.6 — same request shape, different model value:

curl https://inference.dahl.global/v1/chat/completions \
  -H "Authorization: Bearer $DAHL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "moonshotai/Kimi-K2.6",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

DeepSeek V4 Flash — same request shape, different model value:

curl https://inference.dahl.global/v1/chat/completions \
  -H "Authorization: Bearer $DAHL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-ai/DeepSeek-V4-Flash-0731",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Model availability

GET /v1/status reports recent probe results per model (uptime and operational flag). It is public — no API key required.

curl "https://inference.dahl.global/v1/status?window=24h"

Supported window values: 1h, 24h, 7d, 30d (default 24h if omitted). Example response shape:

{
  "models": [
    { "id": "MiniMaxAI/MiniMax-M2.7", "operational": true, "uptime": 98.5 }
  ],
  "window": "24h",
  "updated_at": "2026-07-05T00:00:00Z"
}

Use this when a model appears in GET /v1/models but requests fail or time out — the status page and this endpoint reflect the same probe data.

Streaming

Add "stream": true to the JSON body for token streaming. Dahl forwards streaming responses from the upstream network in OpenAI SSE format (data: {...} lines, ending with data: [DONE]). Streaming uses the same Authorization: Bearer header and model id as non-streaming requests. For long completions, streaming is usually preferable so your client can render tokens as they arrive instead of waiting for the full response body.

curl https://inference.dahl.global/v1/chat/completions \
  -H "Authorization: Bearer $DAHL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMaxAI/MiniMax-M2.7",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": true
  }'

When the upstream includes usage metadata in the stream, successful requests also update your key balance (see Check remaining balance). Manage keys and usage at inference.dahl.global/account.

When a model fails

If you receive a 4xx or 502 mentioning the model:

Call GET /v1/models again
Check GET /v1/status for recent uptime
Switch to an id present in the response
Retry with short backoff if the network reports overload (503 or timeouts)

See Quickstart for a full first-request walkthrough.

Getting started

Guides

Currently available

Default choice

Using a model in chat completions

Model availability

Streaming

When a model fails

​Currently available

​Default choice

​Using a model in chat completions

​Model availability

​Streaming

​When a model fails

Currently available

Default choice

Using a model in chat completions

Model availability

Streaming

When a model fails