Efficient Qwen thinking model for local reasoning, math, and coding agents
Qwen3-Next 80B-A3B (Thinking) by alibaba is offered by 15 providers on this page. Public prices are shown for 14 of them. Its official list price is $0.50 per 1M input tokens and $6.00 per 1M output tokens. The lowest paid channel is NanoGPT at $0.15 / $0.65 per 1M, about 3.5× below the list price. That channel is a third-party gateway, so confirm its availability and rate limits before depending on it.
The context window is 131,072 tokens at the reference host, but hosts report different limits, from 65,536 to 262,144, so the usable window depends on the provider you pick. It supports reasoning, tool use, and structured output. The weights are open, so it can also be self-hosted or served through a gateway of your choice. Providers report the capability flags inconsistently, so verify a specific feature against the host you plan to use. Its training knowledge cuts off at 2025-04.
| Provider | Tier | Input | Output | Cache read | Cache write | Context | Output limit | Status |
|---|---|---|---|---|---|---|---|---|
| NanoGPT | Gateway | $0.15 | $0.65 | $0.075 | — | 256,000 ⚠ | 32,768 | |
| Kilo Gateway | Gateway | $0.15 | $1.20 | — | — | 131,072 | 32,768 | |
| OpenRouter | Gateway | $0.15 | $1.20 | — | — | 262,144 ⚠ | 32,768 | |
| Vercel AI Gateway | Cloud | $0.15 | $1.20 | — | — | 262,144 ⚠ | 262,144 | |
| Merge Gateway | Gateway | $0.15 | $1.20 | — | — | 131,072 | 32,768 | |
| Eden AI | Gateway | $0.15 | $1.20 | — | — | 131,072 | 32,768 | |
| DevPass (LLM Gateway) | Gateway | $0.15 | $1.20 | — | — | 131,072 | 32,768 | |
| LLM Gateway | Gateway | $0.15 | $1.20 | — | — | 131,072 | 32,768 | |
| Alibaba (China)Official | First-party | $0.14 | $1.43 | — | — | 131,072 | 32,768 | |
| NovitaAI | Gateway | $0.15 | $1.50 | — | — | 131,072 | 32,768 | |
Sorted by blended price (input×0.75 + output×0.25) asc. The official channel always shows regardless of rank. Whether a gateway's low price is actually usable can't be verified.
Interleaved thinking (reasoning between tool calls) is declared by 1 of 15 providers.
6 / 15 providers expose no reasoning control (reasoning_options: []).