model apis · chat / llm
Run Qwen3.8 2.4t A95B by Qwen through one OpenAI-compatible endpoint. Pay per use against your GPU.ai balance. No instance to rent, no deployment to manage.
Available nowLive catalog rates, billed per use against your GPU.ai balance. No subscriptions, no minimums.
| Tier | Rate |
|---|---|
| Serverless | Input tokens: $2.00 / 1MCached input tokens: $0.250 / 1MOutput tokens: $6.00 / 1M |
rates are read every few minutes from the live model catalog and metered per request
Any OpenAI SDK works: set the base URL to https://api.gpu.ai/v1 and use your GPU.ai API key.
curl https://api.gpu.ai/v1/chat/completions \
-H "Authorization: Bearer $GPUAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpuai/qwen3.8-2.4t-a95b",
"messages": [{"role": "user", "content": "Hello!"}]
}'from openai import OpenAI
client = OpenAI(
base_url="https://api.gpu.ai/v1",
api_key="YOUR_GPUAI_API_KEY",
)
response = client.chat.completions.create(
model="gpuai/qwen3.8-2.4t-a95b",
messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.gpu.ai/v1",
apiKey: process.env.GPUAI_API_KEY,
});
const response = await client.chat.completions.create({
model: "gpuai/qwen3.8-2.4t-a95b",
messages: [{ role: "user", content: "Hello!" }],
});
console.log(response.choices[0].message.content);| Model id | gpuai/qwen3.8-2.4t-a95b |
| Modality | Chat / LLM |
| Author | Qwen |
| Context window | 1,010,000 tokens |
| Streaming | — |
| Aliases | — |
faq
Qwen3.8 2.4t A95B on GPU.ai currently costs $2.00 / 1M (input tokens), $0.250 / 1M (cached input tokens), $6.00 / 1M (output tokens). There are no subscriptions or minimums. Usage is billed against your GPU.ai balance as you go.
Yes. Point any OpenAI SDK at https://api.gpu.ai/v1 with a GPU.ai API key and request model "gpuai/qwen3.8-2.4t-a95b". No proprietary client needed.
Qwen3.8 2.4t A95B supports a 1,010,000-token context window on GPU.ai.
Chat models bill per input and output token at the listed per-million-token rates, metered per request against your GPU.ai balance. Where a cached-input rate is listed, the part of a prompt the serving partner answers from its prompt cache (reported as prompt_tokens_details.cached_tokens) bills at that lower rate instead of the input rate.