Qwen3 30B A3B Thinking 2507
Qwen · released Jul 29, 2025 · Qwen/Qwen3-30B-A3B-Thinking-2507
- Type
- Open weightsApache License 2.0
- Params
- 30.5B
- Context
- 82K
3B active per word · about 61K words of context
Our take
Written Sep 4, 2026Qwen3 is a thinking-mode model from Alibaba with 30.5 billion total parameters and only 3 billion active per word. It carries a permissive Apache licence and an 81,920-token request limit, though no quality benchmarks have reached our data yet.
Pick this for self-hosted reasoning where parameter efficiency matters — roughly ten times as many total parameters as active ones. Use it when you need an unrestricted Apache licence for commercial deployment or fine-tuning. Skip it if you need measured quality scores, consistent throughput guarantees, or multimodal input.
The case for it
- Only 3 billion active parameters per token from 30.5 billion total — unusually sparse for its size.
- Apache 2.0 licence allows commercial use, fine-tuning and redistribution without restriction.
- Identical per-token pricing across all three tracked offers.
The case against it
- No benchmark scores of any kind in our data — no measured quality or reasoning accuracy.
- Throughput is thin and unexplained: two Alibaba endpoints report 136.5 and 141 tokens per second with no stated cause for the gap.
- Text-only; no image, audio or video input.
How good is it?
We hold no score for this model.
So there is no figure here for everyday use, coding, agent work or writing. That is a gap in our data, not a low score.
Can you run it yourself?
GeForce RTX 4090 · 24 GB
Borderline fit on an estimated size. It leaves 1.6 GB spare on a size we calculated rather than measured, and a 10% error either way would change the answer.
Comfortable fit
GeForce RTX 5090 · 32 GB
Room to spare. 9.6 GB spare means a 10% error in the size would not change the answer.
Apple M1 Pro (16-core GPU) · 32 GB
Room to spare. 2.8 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 4 hours ago — each listing carries its own date.
- per 1M tokens
- $0.20 in / $2.40 out
- Context served
- 82K
- Throughput
- ~125 tok/s
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenRouterOpenRouter's own listing | $0.20 / $2.40checked 4 hours ago | 82K | not measured | Unknown | Unknown | Unknown |
| Alibaba CloudThrough OpenRouter | $0.20 / $2.40checked 4 hours ago | 82K33K max reply | 125 tok/s | No | Yesunknown period | Unknown |
OpenRouterAcross the 2 listings we hold: 1 says it does not train on prompts, 0 say they do and 1 does not say. 0 appear 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 | ✓ | ✓ | ✗ |
| Alibaba CloudThrough OpenRouter | ✓ | ✓ | ✗ |
Tool calling: 2 of 2 listings say yes. JSON output: 2 of 2 listings say yes. Strict schema: 0 of 2 listings say yes, 2 say no.
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.
- No independent board has scored it, so we hold no quality figures at all.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 2 listings does not say whether it trains 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/Qwen3-30B-A3B-Thinking-2507
- Architecture
- Mixture of experts
- Takes in, gives back
- Text in, text out
- Catalogue slug
- qwen-qwen3-30b-a3b-thinking-2507