Qwen3.8-27B

General-purpose open model suitable for coding, agents and structured workloads.

Best for: Coding, agents, structured output

Not yet serving traffic. Benchmarks are published on our serving configuration; general availability has not opened.

Input / 1M: Not yet published
Output / 1M: Not yet published
Context: 128K
Model ID: qwen3.8-27b
Availability: Beta

Pricing

Per-token pricing, metered separately for input, output and cache reads. Input and output rates are in the summary above.

Cache read / 1MNot yet published

Prices are set from the measured cost of serving this model and published before beta access opens. We quote a price we can stand behind rather than a placeholder.

Model origin

Upstream: unsloth/Qwen3.8-27B-GGUF
Base model: Qwen/Qwen3.8-27B
License: apache-2.0

The exact weights, revision hash and serving build are published on the benchmarks page.

Technical details →

API

OpenAI-compatible. Point your client at https://api.crowdrouter.com/v1 and use the model ID qwen3.8-27b. Existing SDKs and clients work unchanged.

Supported

OpenAI-compatible endpoint
Streaming (SSE)
tools / tool_choice
response_format (structured output)
Per-model verified limits

131,072 tokens is the maximum context for a single request. The cache configuration used to serve it is declared on the benchmarks page.

bash

curl https://api.crowdrouter.com/v1/chat/completions \
  -H "Authorization: Bearer $CROWDROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.8-27b",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

How we know

Every figure behind this model traces to a measurement or a named external source. Where something is not yet known, it says so.

Single-request throughputMEASURED
Aggregate throughputMEASURED
Prompt processingMEASURED
Artifact and revisionMEASURED
Serving cost basisDERIVED
Token pricingCURRENT MARKET

Benchmarks and raw data →