InferX Endpoint | Qwen3.6-27B-FP8

Metadata

Name

Qwen3.6-27B-FP8

Provider

Qwen

Parameter Size

27.00B

GPU Count

1

Context Length

262144

Concurrency

2.29x

Cold Start TTFT

1 sec

Recommended Use Cases

Coding assistant, Agent / tool use

Detailed Intro

Qwen3.6-27B-FP8 is a dense, high-performance LLM optimized for efficiency and long-context reasoning.

Strengths
High Density: Superior reasoning and coding logic compared to smaller models.

Context Mastery: Native support for up to 1M tokens, ideal for full-repo analysis.

Efficiency: FP8 quantization halves memory usage with minimal accuracy loss, enabling 300K+ KV cache tokens on a single GPU.

Usage
Complex Coding Agents: Handling massive contexts in tools like OpenCode.

Enterprise RAG: Searching across thousands of documents.

Cost-Effective Scaling: Higher concurrency for serverless platforms like InferX.

Log In To Use This Endpoint

This public page shows the published endpoint metadata and integration shape. Log in to get a tenant-scoped endpoint URL, inference API key, and the interactive playground. Log in

Integration

Start with the shared setup below, then choose the endpoint mode that fits your workload: token-based shared serving or a tenant-scoped dedicated GPU endpoint.

API Key

<INFERENCE_API_KEY>

An inference API key is required for both endpoint modes. Until one is available, the fields below keep the correct request shape and use a placeholder token.

Token-Based

Use the shared OpenAI-compatible endpoint. The base URL is shared across published endpoints, and usage is billed per token.

API Base URL

https://model.inferx.net/endpoints/v1

Model Name

Qwen3.6-27B-FP8

Sample REST Call

curl https://model.inferx.net/endpoints/v1/chat/completions \
  -H "Authorization: Bearer $INFERX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "Qwen3.6-27B-FP8", "messages": [{"role": "user", "content": "Hello"}]}'

Dedicated GPU

Model Spec

{
    "image": "inferx/vllm-openai:v0.20.2-cu129",
    "commands": [
        "--model",
        "Qwen/Qwen3.6-27B-FP8",
        "--trust-remote-code",
        "--gpu-memory-utilization",
        "0.99",
        "--enable-prefix-caching",
        "--enable-auto-tool-choice",
        "--tool-call-parser",
        "qwen3_coder",
        "--uvicorn-log-level",
        "warning",
        "--enable-prefix-caching",
        "--tensor-parallel-size=1",
        "--max-model-len",
        "262144"
    ],
    "resources": {
        "GPU": {
            "Count": 1,
            "vRam": 70000
        }
    },
    "envs": [
        [
            "VLLM_SERVER_DEV_MODE",
            "1"
        ],
        [
            "VLLM_USE_DEEP_GEMM",
            "0"
        ]
    ],
    "policy": {
        "Obj": {
            "nvidia_replica": 0,
            "min_replica": 0,
            "max_replica": 1,
            "standby_per_node": 1,
            "parallel": 50,
            "queue_len": 100,
            "queue_timeout": 30.0,
            "scalein_timeout": 1.0,
            "scaleout_policy": {
                "WaitQueueRatio": {
                    "wait_ratio": 0.1
                }
            },
            "runtime_config": {
                "graph_sync": false
            }
        }
    },
    "sample_query": {
        "apiType": "text2text",
        "raw_query": false,
        "path": "v1/chat/completions",
        "prompt": "write a quicksort with python3",
        "prompts": [],
        "dataUrl": "",
        "max_tokens": 0,
        "body": {
            "max_tokens": 1000,
            "messages": [
                {
                    "content": "You are a helpful assistant.",
                    "role": "system"
                },
                {
                    "content": "write a quicksort with python3",
                    "role": "user"
                }
            ],
            "model": "Qwen/Qwen3.6-27B-FP8",
            "stream": true,
            "temperature": 0
        },
        "loadingTimeout": 90
    }
}