Qwen3.5-122B-A10B
Qwen · released Feb 24, 2026 · Qwen/Qwen3.5-122B-A10B
- Type
- Open weightsApache License 2.0
- Params
- 125B
- Context
- 262K
10B active per word · about 197K words of context
Our take
Written Aug 2, 2026Qwen3.5-122B-A10B is a large downloadable model from Alibaba that uses a mixture-of-experts design, activating only 10 billion of its 125.1 billion parameters for each token. It scores highest on coding benchmarks among all scores we hold for this model, and carries a permissive Apache licence.
Choose this for frontier coding workloads where measured ability matters — its Arena Coding score is the highest we hold for this model. Use it for long-context multimodal pipelines needing up to 262,144 tokens, or for cost-sensitive deployment under a permissive licence. Skip it if creative writing or web development coding is your main need, where it lags its own general coding score by a wide margin.
The case for it
- Top-tier measured coding ability: Arena Coding Elo 1458.7997, 42.9 points above its overall text Elo.
- Dramatic efficiency from its mixture-of-experts design: only 10 billion active parameters per token from 125.1 billion total.
- Apache License 2.0 allows commercial use, fine-tuning and redistribution.
- Wide provider choice with competitive pricing: nine offers across five providers.
The case against it
- Creative writing lags other capabilities: 91.7 points below its coding score and 49.8 points below its overall text score.
- Web development coding trails general coding: 99.4 points below its Arena Coding score.
- Throughput is highly variable across providers, from 2 to 64 tokens per second, with no throughput disclosed for three of nine offers.
How good is it?
IntelligencePuzzles, maths, exam questions
Arena Text (overall)62nd of 143 · 1416.9
CodingWriting and fixing code on its own
Arena Coding66th of 143 · 1459
AgenticPlanning, calling tools, staying on task
Nobody we watch has scored Qwen3.5-122B-A10B 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 Qwen3.5-122B-A10B 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 M1 Ultra (64-core GPU) · 128 GB
Room to spare. 13.4 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 9 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
- $0.26 in / $2.08 out
- Context served
- 262K
- Throughput
- ~54 tok/s
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| Alibaba Cloudfp8 | $0.26 / $2.08 | 262K | 169 tok/s | No | Yesunknown period | Unknown |
| Alibaba Cloud | $0.26 / $2.08 | 262K | 54 tok/s | No | Yesunknown period | Unknown |
| SiliconFlowfp8 | $0.26 / $2.08 | 262K | 77 tok/s | No | No | Confirmed |
| AtlasCloudfp8 | $0.30 / $2.40 | 262K | 140 tok/s | No | Yesunknown period | Unknown |
| DeepInfrafp4 | $0.29 / $2.40 | 262K | not measured | Unknown | Unknown | Unknown |
| DeepInfrafp4 | $0.29 / $2.40 | 262K | 152 tok/s | No | No | Confirmed |
| Novita AIbf16 | $0.40 / $3.20 | 262K | 132 tok/s | No | No | Confirmed |
| Novita AI | $0.40 / $3.20 | 262K | not measured | Unknown | Unknown | Unknown |
| OpenRouter | $0.40 / $3.20 | 262K | not measured | Unknown | Unknown | Unknown |
DeepInfra
Novita AI
OpenRouterAcross the 9 listings we hold: 6 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 |
|---|---|---|---|
| Alibaba Cloudfp8 | ✓ | ✓ | ✓ |
| Alibaba Cloud | ✓ | ✓ | ✓ |
| SiliconFlowfp8 | ✗ | ✓ | ✓ |
| AtlasCloudfp8 | ✓ | ✓ | ✓ |
| DeepInfrafp4 | |||
| DeepInfrafp4 | ✓ | ✗ | ✗ |
| Novita AIbf16 | ✓ | ✓ | ✗ |
| Novita AI | |||
| OpenRouter | ✓ | ✓ | ✓ |
Tool calling: 6 of 9 listings say yes, 1 says no, 2 publish no parameter list. JSON output: 6 of 9 listings say yes, 1 says no, 2 publish no parameter list. Strict schema: 5 of 9 listings say yes, 2 say no, 2 publish no parameter list.
Models people weigh against Qwen3.5-122B-A10B
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 9 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 9 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/Qwen3.5-122B-A10B
- Architecture
- Mixture of experts
- Modality record
- text+image+video->text
- Catalogue slug
- qwen-qwen3-5-122b-a10b