Qwen3 Next 80B A3B Thinking
Qwen · released Sep 9, 2025 · Qwen/Qwen3-Next-80B-A3B-Thinking
- Type
- Open weightsApache License 2.0
- Params
- 81.3B
- Context
- 262K
3B active per word · about 197K words of context
Our take
Written Sep 2, 2026Qwen3 Next is a thinking-mode text model with an Apache licence and a 262,144-token request limit. It uses a mixture-of-experts design where only 3 billion of its 81.3 billion parameters activate per token, making it unusually parameter-efficient for its scale.
Choose this for long-context text work needing a quarter-million tokens of history, or for Apache-licensed commercial deployment where thinking-mode output is useful. It suits budget-conscious workloads with uniform low input pricing across all tracked offers. Skip it if you need image, video or audio input, or if creative writing quality is your priority — that is its weakest measured dimension.
The case for it
- Extreme parameter efficiency: 81.3 billion total with only 3 billion active per token, a 27:1 compression ratio.
- Coding score 52.6 points above its general text ranking on the same leaderboard.
- All six tracked offers charge the same low input rate, with five matching on output too.
- Primary vendor throughput of 200 tokens per second, roughly four times the slowest tracked option.
The case against it
- Creative writing is its weakest measured dimension, 62.7 points below its coding score.
- One provider charges a premium on output versus the other five tracked offers.
- Throughput unverified for two of the six offers.
How good is it?
EverydayGeneral questions and everyday reasoning
Arena Text (overall)113th of 168 · 1369
CodingWriting and fixing code on its own
Arena Coding114th of 168 · 1422
Arena Coding is the only board that has scored it for this.
AgenticPlanning, calling tools, staying on task
Not yet scored on Arena Agent.
WritingDrafting and rewriting prose
Arena Creative Writing115th of 168 · 1326
Arena Creative Writing is the only board that has scored it for this.
These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done.
Every published score for this model6 scoresEvery figure we hold, from 6 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
GeForce RTX 4090 · 24 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Apple M1 Pro (16-core GPU) · 32 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Comfortable fit
Apple M2 Max (38-core GPU) · 96 GB
Room to spare. 17.8 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked between 4 hours and 21 days ago — each listing carries its own date.
- per 1M tokens
- $0.15 in / $1.20 out
- Context served
- 262K
- Throughput
- ~46 tok/s
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenRouterOpenRouter's own listing | $0.15 / $1.20checked 4 hours ago | 262K | not measured | Unknown | Unknown | Unknown |
| Google Vertex AIglobalThrough OpenRouter | $0.15 / $1.20checked 4 hours ago | 262K236K max reply | 46 tok/s | No | No | Confirmed |
| Alibaba CloudThrough OpenRouter | $0.15 / $1.20checked 4 hours ago | 131K33K max reply | 232 tok/s | No | Yesunknown period | Unknown |
| Novita AIDirect | $0.15 / $1.50checked 21 days ago | 131K | not measured | Unknown | Unknown | Unknown |
OpenRouter
Novita AIAcross the 4 listings we hold: 2 say they do not train on prompts, 0 say they do and 2 do not say. 1 appears in the zero-retention registry we check; the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| OpenRouterOpenRouter's own listing | ✓ | ✓ | ✓ |
| Google Vertex AIglobalThrough OpenRouter | ✓ | ✓ | ✓ |
| Alibaba CloudThrough OpenRouter | ✓ | ✓ | ✗ |
| Novita AIDirect |
Tool calling: 3 of 4 listings say yes, 1 publishes no parameter list. JSON output: 3 of 4 listings say yes, 1 publishes no parameter list. Strict schema: 2 of 4 listings say yes, 1 says no, 1 publishes no parameter list.
When we formed this view
Recent changes
What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- 1 of 4 listings publishes no parameter list, so what its API accepts is unknown to us.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 2 of 4 listings do not say whether they train on prompts.
- We hold no cached-input rate for any of its listings.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsApache License 2.0, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Apache License 2.0
Fully permissive: commercial use, redistribution, and derivatives allowed. Requires attribution and a copy of the license. Includes an express patent grant.
Identifiers
- Hugging Face
- Qwen/Qwen3-Next-80B-A3B-Thinking
- Architecture
- Mixture of experts
- Takes in, gives back
- Text in, text out
- Catalogue slug
- qwen-qwen3-next-80b-a3b-thinking