Qwen3 30B A3B Instruct 2507
Qwen · released Jul 28, 2025 · Qwen/Qwen3-30B-A3B-Instruct-2507
- Type
- Open weightsApache License 2.0
- Params
- 30.5B
- Context
- 262K
3B active per word · about 197K words of context
Our take
Written Aug 2, 2026Qwen 3 is a text-only model from Alibaba with a mixture-of-experts design that keeps only 3 billion parameters active for each word while drawing on 30.5 billion total. It scores highest on coding leaderboards and is available from ten hosts, with entry-level pricing well below the vendor's own rate.
Choose this for coding workloads where measured leaderboard performance matters, or for cost-sensitive text inference with long contexts up to 262,144 tokens. It suits local or edge deployment where low active-parameter counts reduce memory pressure. Skip it if you need image, video or audio support, if creative writing quality is central, or if you want the cheapest option and can only access first-party pricing.
The case for it
- Only 3 billion active parameters from 30.5 billion total — roughly a ten-to-one sparsity ratio that eases memory and compute demands.
- Coding is its standout measured skill, with an Arena Coding Elo 118.4 points above its creative-writing score.
- Ten current hosted offers, with entry-level pricing roughly 2.7 times cheaper than the first-party rate.
- Apache 2.0 licence allows commercial use, fine-tuning and redistribution.
The case against it
- Creative writing is its weakest measured category, sitting 62.2 points below its overall text score.
- First-party pricing from Alibaba is among the most expensive options tracked.
- Text-to-text only; no image, video or audio support.
How good is it?
IntelligencePuzzles, maths, exam questions
Arena Text (overall)86th of 143 · 1383.1
CodingWriting and fixing code on its own
Arena Coding78th of 143 · 1439.3
Arena Coding is the only board that has scored it for this.
AgenticPlanning, calling tools, staying on task
Nobody we watch has scored Qwen3 30B A3B Instruct 2507 for this. We would take the rating from Arena Agent (IPS).
WritingWe do not rate this
Two boards come close and neither tests writing: Arena Creative Writing asks people which of two replies they prefer, and LiveBench Language tests whether a model understood a passage. So we show where Qwen3 30B A3B Instruct 2507 placed and give it no mark out of five.
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, which is why they get no rating.
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?
- Fits in memory
- weights load entirely on the card
- Spills to system RAM
- some weights offload; much slower
- Too large
- will not load even with offload
- est
- size is calculated; the verdict could change by 10%
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.
All 3 sizes here are calculated, not measured. We hold no measured file for this model, so each size comes from the parameter count and every verdict above inherits that uncertainty.
Check against your own machine → · All 71 devices, with every size →
Or rent it from someone else
Cheapest of 12 live listings. Picked at the widest standard context we hold, within one quantisation slice, so the numbers beside it are a price one host actually charges.
- per 1M tokens
- $0.048 in / $0.19 out
- Context served
- 262K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| StreamLake | $0.048 / $0.19 | 128K | 20 tok/s | No | Yesunknown period | Unknown |
| OpenRouter | $0.048 / $0.19 | 262K | not measured | Unknown | Unknown | Unknown |
| Nebius AI Studiofp8 | $0.10 / $0.30 | 262K | 38 tok/s | No | No | Confirmed |
| SiliconFlowfp8 | $0.090 / $0.30 | 262K | 26 tok/s | No | No | Confirmed |
| CoreWeavebf16 | $0.10 / $0.30 | 262K | 33 tok/s | No | No | Confirmed |
| Novita AI | $0.090 / $0.45 | 41K | not measured | Unknown | Unknown | Unknown |
| DeepInfrafp8 | $0.12 / $0.50 | 41K | 63 tok/s | No | No | Confirmed |
| DeepInfrafp8 | $0.12 / $0.50 | 41K | not measured | Unknown | Unknown | Unknown |
| Alibaba Cloud | $0.13 / $0.52 | 131K | 68 tok/s | No | Yesunknown period | Unknown |
| NextBitfp8 | $0.12 / $0.52 | 33K | 5 tok/s | No | No | Unknown |
| Alibaba Cloudfp8 | $0.13 / $0.52 | 131K | 95 tok/s | No | Yesunknown period | Unknown |
| Phala | $0.15 / $0.55 | 262K | 69 tok/s | No | No | Unknown |
Nebius AI Studio
Novita AI
DeepInfraAcross the 12 listings we hold: 9 say they do not train on prompts, 0 say they do and 3 do not say. 4 appear in the zero-retention registry we check; the rest are unknown to us rather than confirmed either way.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
- ✓
- Supported
- ✗
- Not supported
- Not published
- host gave no parameter list
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| StreamLake | ✓ | ✓ | ✓ |
| OpenRouter | ✓ | ✓ | ✓ |
| Nebius AI Studiofp8 | ✓ | ✓ | ✓ |
| SiliconFlowfp8 | ✓ | ✓ | ✓ |
| CoreWeavebf16 | ✓ | ✓ | ✓ |
| Novita AI | |||
| DeepInfrafp8 | ✓ | ✓ | ✗ |
| DeepInfrafp8 | |||
| Alibaba Cloud | ✓ | ✓ | ✗ |
| NextBitfp8 | ✗ | ✓ | ✓ |
| Alibaba Cloudfp8 | ✓ | ✓ | ✗ |
| Phala | ✓ | ✗ | ✓ |
Tool calling: 9 of 12 listings say yes, 1 says no, 2 publish no parameter list. JSON output: 9 of 12 listings say yes, 1 says no, 2 publish no parameter list. Strict schema: 7 of 12 listings say yes, 3 say no, 2 publish no parameter list.
Models people weigh against Qwen3 30B A3B Instruct 2507
When we formed this view
Dates behind this page
Prices last checked 37h ago
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.
- 2 of 12 listings publish no parameter list, so what their API accepts is unknown to us.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 3 of 12 listings do not say whether they train on prompts.
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-Instruct-2507
- Architecture
- Mixture of experts
- Modality record
- text->text
- Catalogue slug
- qwen-qwen3-30b-a3b-instruct-2507