- Type
- Open weightsApache License 2.0
- Params
- 32.8B
- Context
- 131K
about 98K words of context
Our take
Written Aug 2, 2026Qwen3 is a 32.8-billion-parameter text model released in 2025 with a permissive Apache licence. It scores highest on coding tasks among its measured skills, and hosts offer a wide range of price and speed trade-offs.
Pick this for budget-conscious hosted deployment where the cheapest tier in its family keeps costs low, or for latency-sensitive workloads where one host delivers 354 tokens per second. It is also a sound choice for self-hosting or commercial products thanks to its Apache 2.0 licence. Skip it if creative writing quality is your main need, or if you need verified efficiency claims about active parameters.
The case for it
- Strongest measured skill is coding, with an Arena Coding Elo 102.6 points above its creative-writing score.
- Apache 2.0 licence permits commercial use, fine-tuning and redistribution.
- Eight tracked offers span a wide speed range, from 20 to 354 tokens per second, with corresponding price trade-offs.
The case against it
- Creative writing is its weakest measured skill, 102.6 points below its coding score and 62.8 points below its hard-prompts score.
- Active parameter count is undisclosed, so efficiency claims remain unverified.
How good is it?
IntelligencePuzzles, maths, exam questions
Arena Text (overall)102nd of 143 · 1347.2
CodingWriting and fixing code on its own
Arena Coding99th of 143 · 1407.2
Arena Coding is the only board that has scored it for this.
AgenticPlanning, calling tools, staying on task
Nobody we watch has scored Qwen3 32B 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 32B 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
Loads, but do not expect an assistant. Some of the weights sit in ordinary system memory, which is far slower than the card.
Comfortable fit
GeForce RTX 5090 · 32 GB
Room to spare. 7.8 GB spare means a 10% error in the size would not change the answer.
Apple M1 Pro (16-core GPU) · 32 GB
Borderline fit on an estimated size. It leaves 1 GB spare on a size we calculated rather than measured, and a 10% error either way would 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 8 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
- $0.080 in / $0.28 out
- Context served
- 131K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenRouter | $0.080 / $0.28 | 131K | not measured | Unknown | Unknown | Unknown |
| DeepInfrafp8 | $0.080 / $0.28 | 41K | 31 tok/s | No | No | Confirmed |
| DeepInfrafp8 | $0.080 / $0.28 | 41K | not measured | Unknown | Unknown | Unknown |
| Nebius AI Studiofp8 | $0.10 / $0.30 | 41K | 23 tok/s | No | No | Unknown |
| Nebius AI Studiobasefp8 | $0.10 / $0.30 | 41K | 25 tok/s | No | No | Unknown |
| Novita AI | $0.10 / $0.45 | 41K | not measured | Unknown | Unknown | Unknown |
| SiliconFlowfp8 | $0.14 / $0.57 | 131K | 16 tok/s | No | No | Confirmed |
| Groq | $0.29 / $0.59 | 131K | 361 tok/s | No | No | Confirmed |
DeepInfra
Nebius AI Studio
Novita AI
GroqAcross the 8 listings we hold: 5 say they do not train on prompts, 0 say they do and 3 do not say. 3 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 |
|---|---|---|---|
| OpenRouter | ✓ | ✓ | ✓ |
| DeepInfrafp8 | ✓ | ✓ | ✓ |
| DeepInfrafp8 | |||
| Nebius AI Studiofp8 | ✓ | ✓ | ✓ |
| Nebius AI Studiobase · fp8 | ✓ | ✓ | ✓ |
| Novita AI | |||
| SiliconFlowfp8 | ✓ | ✓ | ✓ |
| Groq | ✗ | ✓ | ✗ |
Tool calling: 5 of 8 listings say yes, 1 says no, 2 publish no parameter list. JSON output: 6 of 8 listings say yes, 2 publish no parameter list. Strict schema: 5 of 8 listings say yes, 1 says no, 2 publish no parameter list.
Models people weigh against Qwen3 32B
When we formed this view
Dates behind this page
Prices last checked 4d 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 8 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 8 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/Qwen3-32B
- Architecture
- Dense
- Modality record
- text->text
- Catalogue slug
- qwen-qwen3-32b