# Dedicated LLM inference

Planning tools: [LLM inference cost and capacity calculator](/dedicated-inference/calculator) and [dedicated inference benchmarks explained](/benchmarks/dedicated-inference).

Reserved B200 or B300 capacity behind an isolated, OpenAI compatible endpoint. Morph runs the stack. You pay for GPU time.

## Deploy from a coding agent

Claude Code or Codex can choose, buy, and monitor an endpoint through the Morph CLI. Purchases require your approval.

```bash
morph dedicated models --json
morph dedicated create deepseek-v4-flash --balanced --yes
morph dedicated status <endpoint-id> --json
morph dedicated logs <endpoint-id> --follow
```

## When dedicated fits

Choose dedicated for steady traffic, a proven model, and workloads constrained by shared limits. Choose serverless while demand is uncertain.

Common workloads include coding agents, production chat, document processing, and multi step pipelines.

## Dedicated versus serverless

| Decision | Dedicated | Serverless |
| --- | --- | --- |
| Traffic | Steady | Variable |
| Billing | GPU time | Tokens |
| Capacity | Reserved | Shared |
| Limits | Purchased capacity | Account limits |
| Best for | Control | No idle cost |

## Models

| Model | Best for | Context |
| --- | --- | --- |
| DeepSeek V4 Flash 0731 | Coding, reasoning, tools | 1M |
| Qwen 3.6 27B | Fast chat, extraction, multimodal | 131K |
| Gemma 4 31B | Reasoning, tools, vision | 175K |
| MiniMax M3 428B | Agents, software, reasoning | 256K |
| GLM 5.3 Flash | Multimodal coding and tools | 1M |
| GLM 5.2 744B | Complex coding and tools | 1M |
| Kimi K3 2.8T | Frontier agents, vision, long runs | 1M |

Test with your own traces before buying. Prices are at https://www.morphllm.com/dedicated-inference#pricing.

Monthly price equals the GPU rate times reserved GPUs times 730 hours. Each month is invoiced upfront, including idle capacity. Tokens are not billed.

## OpenAI compatible API

Keep the OpenAI SDK. Change the base URL and model.

```python
from openai import OpenAI
import os

client = OpenAI(
    api_key=os.environ["MORPH_API_KEY"],
    base_url="https://your-endpoint.morphllm.com/v1",
)

response = client.chat.completions.create(
    model="your-dedicated-model",
    messages=[{"role": "user", "content": "Hello"}],
)
```

## Operations

Morph handles provisioning, serving, monitoring, incidents, tuning, and migration. The dashboard shows traffic, tokens, cache, latency, status, and capacity. The CLI streams metadata logs.

Service includes a 99.9 percent monthly availability SLA, Slack support, and customer success. Your agreement defines measurement and remedies.

The initial term is 90 days, then monthly with 30 days notice. Provisioning starts after payment. If activation exceeds two hours, Morph cancels and refunds the order.

## FAQ

### Is hardware exclusive?

No. Your endpoint, credentials, and capacity are isolated. Idle hardware may serve other traffic.

### Does Morph retain prompts or responses?

No. Morph stores only operational and billing metadata.

### Are tokens capped?

No. You buy GPU time, not tokens.

### Can an agent manage the endpoint?

Yes. Agents can buy, track, list, and inspect endpoints through the Morph CLI after user approval.

### How do I monitor it?

Use the dashboard for metrics and the CLI for status and metadata logs.