Qwen3 Next 80B A3B Thinking
Qwen3-Next uses a highly sparse MoE design: 80B total parameters, but only ~3B activated per inference step.
text → text · made by Alibaba
Sold by 11 ways
| Seller | Lane | Rate |
|---|---|---|
| AWS Bedrock 0 ways | $0.15per Mtok in$1.2per Mtok out | |
| AWS Bedrock cloudnot on the seller's list today | standard | $0.15per Mtok in$1.2per Mtok outraw.githubusercontent.com · read 2026-08-27 |
| Nebius Token Factory 0 ways | $0.15per Mtok in$1.2per Mtok out | |
| Nebius Token Factory aggregatornot on the seller's list today | standard | $0.15per Mtok in$1.2per Mtok outtokenfactory.nebius.com · read 2026-08-25 |
| Nous Research aggregator | standard | $0.15per Mtok in$1.2per Mtok outinference-api.nousresearch.com · read 2026-09-12 |
| Novita AI 0 ways | $0.15per Mtok in$1.5per Mtok out | |
| Novita AI aggregatornot on the seller's list today | standard | $0.15per Mtok in$1.5per Mtok outapi.novita.ai · read 2026-08-25 |
| OpenRouter aggregator | standard | $0.15per Mtok in$1.2per Mtok outopenrouter.ai · read 2026-08-24 |
| Requesty aggregator | nebius | $0.15per Mtok in$1.2per Mtok out$0.15per Mtok cachedrouter.requesty.ai · read 2026-08-28 |
| Together AI aggregator | standard | $0.15per Mtok in$1.5per Mtok outraw.githubusercontent.com · read 2026-11-01 |
| Vercel AI Gateway aggregator | standard | $0.15per Mtok in$1.2per Mtok outai-gateway.vercel.sh · read 2026-08-25 |
| Hugging Face Inference Providers 0 ways | $0.3per Mtok in$2per Mtok out | |
| Hugging Face Inference Providers aggregatornot on the seller's list today | standard | $0.3per Mtok in$2per Mtok outmodels.dev · read 2026-10-22 |
| Fireworks AI aggregator | standard | $0.9per Mtok in$0.9per Mtok outraw.githubusercontent.com · read 2026-11-01 |
| PPIO 0 ways | $1per Mtok in$10per Mtok out | |
| PPIO aggregatornot on the seller's list today | standard | $1per Mtok in$10per Mtok outapi.ppinfra.com · read 2026-09-10 |
Measured 1 standing
| Place | Board | Metric | Score |
|---|---|---|---|
| 52ndof 111 | Vectara Hallucination Leaderboard | Hallucination Rate | 9.3 |
About
Qwen3 Next 80B A3B Thinking — a text model from Alibaba, sold by 6 companies from $0.15 in and $1.2 out per million tokens, placed 52nd of 111 on Vectara Hallucination Leaderboard.
It takes text and returns text, with a context window of 262,144 tokens. It was published in September 2025, trained on material up to April 2025. Its sellers say it can reason step by step and call a tool. The catalogue files it under extract and reasoning. Six companies sell it. The cheapest is $0.15 in and $1.2 out per million tokens at Nous Research. Beside the standard rate there is a separately routed lane. It stands 52nd of 111 on Vectara Hallucination Leaderboard.
Every current figure
- Maker
- Alibaba
- Register
- model
- Takes
- text
- Returns
- text
- Context
- 262,144 tokens
- Longest answer
- 131,072 tokens
- Published
- September 2025
- Knowledge to
- April 2025
- Parameters
- 81 billion
- Licence
- not read
- Sellers
- 6
- Maker's own price
- not read
- Price
- $0.15 in and $1.2 out per million tokens — Nous Research
- Boards
- 1
- Best place
- 52nd of 111 — Vectara Hallucination Leaderboard
Known as 3 names
Qwen/Qwen3-Next-80B-A3B-Thinkingnebius/qwen/qwen3-next-80b-a3b-thinkingalibaba/qwen3-next-80b-a3b-thinking