Step 1
Get an API key
Register at TokenRa, open the dashboard, and create an API key. Store it server-side.
Qwen 4 Max API · Alibaba Qwen team · via TokenRa
Qwen 4 Max is live on TokenRa. Model ID: qwen4-max. OpenAI-compatible endpoint, no subscription, billed per token.
Qwen 4 Max is live on TokenRa. Everything in the table below is confirmed.
Qwen 4 Max is confirmed and live. Plus, Flash, and 27B have not been officially announced; those columns stay empty until Qwen publishes them.
| Tier | Role | Context window | Input / 1M | Output / 1M | Cache read / 1M | Status |
|---|---|---|---|---|---|---|
| Qwen 4 Max | Flagship API model | 1,000,000 tokens Confirmed | $1.34 | $4.03 | $0.095 | Live on TokenRa |
| Qwen 4 Plus | Middle tier | Not announced | Not announced | Not announced | Not announced | Not announced |
| Qwen 4 Flash | Low-latency, high-volume | Not announced | Not announced | Not announced | Not announced | Not announced |
| Qwen 4 27B | Open-weight release | Not announced | Not announced | Not announced | Not announced | Not announced |
Qwen 4 Max is billed per token on TokenRa. No subscription. Your account needs a positive balance before sending requests; the minimum top-up is $10.
| Model | Context window | Input / 1M | Output / 1M | Cache read / 1M |
|---|---|---|---|---|
| Qwen 4 Max Live | 1,000,000 tokens Confirmed | $1.34 | $4.03 | $0.095 |
| Qwen3.8-Omni-Flash (compare) | 1,000,000 tokens | $0.23 | $0.76 | $0.03 |
Qwen 4 Max is priced at 30% off Alibaba's official rate. Both models share the same 1,000,000-token context window. Qwen 4 Max costs 5.8× more per input token, 5.3× more per output token, and 3.2× more per cached read than Qwen3.8-Omni-Flash. Confirm current rates in the console before production use.
Qwen 4 Max and Qwen3.8-Max are two separate models with separate model IDs and separate pricing. They are not the same model under two names. Qwen3.8-Max is an earlier Qwen flagship; Qwen 4 Max (qwen4-max) is the newer one, now live on TokenRa.
Step 1
Register at TokenRa, open the dashboard, and create an API key. Store it server-side.
Step 2
Use the model ID qwen4-max in any OpenAI-compatible client. No other configuration needed.
Step 3
Point the base URL at tokenra.io/v1 and pass your key as the Authorization Bearer token.
curl https://tokenra.io/v1/chat/completions \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"qwen4-max","messages":[{"role":"user","content":"Summarize the key decisions in this transcript."}]}'
Yes. Qwen 4 Max is live on TokenRa. The model ID is qwen4-max, callable through the OpenAI-compatible endpoint.
Qwen 4 Max costs $1.34 per 1M input tokens, $4.03 per 1M output tokens, and $0.095 per 1M cached reads on TokenRa — 30% off Alibaba's official rate. The minimum top-up is $10. There is no subscription fee — usage is billed per token.
Qwen 4 Max uses the OpenAI-compatible endpoint. Set the base URL to https://tokenra.io/v1, add your TokenRa API key as the Authorization Bearer token, and set model to qwen4-max.
Qwen 4 Max is developed and operated by Alibaba's Qwen team. TokenRa provides API access as a routing gateway and is not the model developer or owner.
Yes. Qwen 4 Max has a 1,000,000-token context window, shared between input and output. Non-thinking mode: max input 991,808 tokens, max output 131,072 tokens. Thinking mode: max input 983,616 tokens, max output 262,144 tokens (thinking content + final answer combined).
Qwen 4 Max costs 5.8× more per input token ($1.34 vs $0.23), 5.3× more per output token ($4.03 vs $0.76), and 3.2× more per cached read ($0.095 vs $0.03) than Qwen3.8-Omni-Flash. Both share the same 1,000,000-token context window. Qwen3.8-Omni-Flash is the lower-cost omni-modal tier.
Qwen 4 Max (qwen4-max) and Qwen3.8-Max are two separate model IDs with their own pricing. They are not the same model under two names.
Qwen 4 Max is developed by Alibaba's Qwen team. Provider and related names belong to their respective rights holders. TokenRa provides API access as a routing gateway and is not the model developer or owner. This page is maintained by TokenRa and reflects public information as of the date shown above.