Qwen3 235B A22B Instruct 2507
Qwen · released Jul 21, 2025 · Qwen/Qwen3-235B-A22B-Instruct-2507
- Type
- Open weightsApache License 2.0
- Params
- 235B
- Context
- 262K
22B active per word · about 197K words of context
Our take
Written Aug 2, 2026Qwen3 is a large mixture-of-experts model from Alibaba with 235.1 billion total parameters and 22 billion active per token, released in July 2025 under a permissive Apache licence. It offers strong measured coding performance and a 262,144-token request limit, though it handles text only with no image, video or audio support.
Choose this for open-weights coding work where measured quality matters — it scores well on LiveCodeBench and the arena coding leaderboard. It is also a sensible pick for long-context text tasks or when you want a permissive licence with commercial freedom. Skip it if you need multimodal input, if creative writing quality is your priority, or if you want the cheapest host and are not prepared to shop between providers.
The case for it
- Strong measured coding performance for a downloadable model: 80.4% on LiveCodeBench and arena coding scores up to 1472.5.
- Efficient mixture-of-experts design with 235.1 billion total parameters but only 22 billion active per token.
- Apache 2.0 licence allows commercial use, fine-tuning and redistribution.
- Ten tracked providers, with several offering throughput above 34 tokens per second and one reaching 46.
The case against it
- Creative writing is its weakest measured category, trailing its own coding peak by more than 35 points on the arena leaderboard.
- Text-only: no image, video or audio input or output, unlike many frontier alternatives.
- Provider pricing varies sharply — the most expensive tracked host charges more than double the cheapest input rate.
How good is it?
IntelligencePuzzles, maths, exam questions
Arena Text (overall)58th of 143 · 1422.9
Also on this board: 1402.9 via Thinking (Aug 2, 2026). Read the pair, not the higher one.
CodingWriting and fixing code on its own
Arena Coding55th of 143 · 1472.5
Also on this board: 1445.8 via Thinking (Aug 2, 2026). Read the pair, not the higher one.
AgenticPlanning, calling tools, staying on task
Nobody we watch has scored Qwen3 235B A22B 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 235B A22B 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 model7 scoresEvery figure we hold, from 7 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
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 M3 Ultra (80-core GPU) · 512 GB
Room to spare. 229.7 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 15 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
- $0.15 in / $0.60 out
- Context served
- 262K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| DeepInfrafp8 | $0.090 / $0.55 | 262K | not measured | Unknown | Unknown | Unknown |
| DeepInfrafp8 | $0.090 / $0.55 | 262K | 15 tok/s | No | No | Confirmed |
| Novita AIfp8 | $0.090 / $0.58 | 131K | 32 tok/s | No | No | Confirmed |
| Novita AI | $0.090 / $0.58 | 131K | not measured | Unknown | Unknown | Unknown |
| OpenRouter | $0.15 / $0.60 | 262K | not measured | Unknown | Unknown | Unknown |
| Alibaba Cloud | $0.15 / $0.60 | 131K | 42 tok/s | No | Yesunknown period | Unknown |
| Alibaba Cloudfp8 | $0.15 / $0.60 | 131K | 39 tok/s | No | Yesunknown period | Unknown |
| Nebius AI Studiofp8 | $0.20 / $0.60 | 262K | 27 tok/s | No | No | Confirmed |
| Venice AIfp8 | $0.15 / $0.75 | 128K | 15 tok/s | No | No | Confirmed |
| Crusoebf16 | $0.22 / $0.80 | 262K | 24 tok/s | No | No | Confirmed |
| Parasailfp8 | $0.14 / $0.80 | 131K | 26 tok/s | No | No | Confirmed |
| Friendli | $0.20 / $0.80 | 262K | 23 tok/s | No | Yesunknown period | Unknown |
| StreamLake | $0.21 / $0.84 | 128K | 45 tok/s | No | Yesunknown period | Unknown |
| AtlasCloudfp8 | $0.20 / $0.88 | 131K | 26 tok/s | No | Yesunknown period | Unknown |
| Google Vertex AIus-south1 | $0.22 / $0.88 | 262K | 37 tok/s | No | No | Confirmed |
DeepInfra
Novita AI
Nebius AI Studio
Google Vertex AIAcross the 15 listings we hold: 12 say they do not train on prompts, 0 say they do and 3 do not say. 7 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 |
|---|---|---|---|
| DeepInfrafp8 | |||
| DeepInfrafp8 | ✓ | ✓ | ✓ |
| Novita AIfp8 | ✓ | ✓ | ✓ |
| Novita AI | |||
| OpenRouter | ✓ | ✓ | ✓ |
| Alibaba Cloud | ✓ | ✓ | ✗ |
| Alibaba Cloudfp8 | ✓ | ✓ | ✗ |
| Nebius AI Studiofp8 | ✓ | ✓ | ✓ |
| Venice AIfp8 | ✓ | ✓ | ✓ |
| Crusoebf16 | ✓ | ✓ | ✓ |
| Parasailfp8 | ✓ | ✓ | ✓ |
| Friendli | ✓ | ✗ | ✓ |
| StreamLake | ✓ | ✓ | ✓ |
| AtlasCloudfp8 | ✓ | ✓ | ✓ |
| Google Vertex AIus-south1 | ✓ | ✓ | ✓ |
Tool calling: 13 of 15 listings say yes, 2 publish no parameter list. JSON output: 12 of 15 listings say yes, 1 says no, 2 publish no parameter list. Strict schema: 11 of 15 listings say yes, 2 say no, 2 publish no parameter list.
Models people weigh against Qwen3 235B A22B Instruct 2507
When we formed this view
Dates behind this page
Prices last checked 34h 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 15 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 15 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/Qwen3-235B-A22B-Instruct-2507
- Architecture
- Mixture of experts
- Modality record
- text->text
- Catalogue slug
- qwen-qwen3-235b-a22b-instruct-2507