Qwen3.6-35B-A3B
- In Input modalities: TextImageVideo Out Output modalities: Text
- Reasoning
- Tool use
- Structured output
- Open weights
Prices
6 offerings · prices as listedqwen3.6-35b-a3b| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | $0.248 | $1.485 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Nov 2, 2026)
qwen3.6-35b-a3b| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | $0.375 | $2.25 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Nov 2, 2026)
qwen3.6-35b-a3b| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | ¥1.80 | ¥10.80 |
Prices from 阿里云百炼：模型价格 (opens in a new tab) (checked Nov 2, 2026)
qwen3.6-35b-a3b| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 21, 2026 | $0.248 | $1.485 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Nov 2, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- Reasoning can be off
- —
- Max thinking budget
- 131.07K
- Built-in tools
- Web search
- Licence
- apache-2.0
- Family
- qwen
- Model ID
Not stated on Alibaba Qwen’s pages: max input, reasoning efforts, knowledge cutoff, aliases.
Run locally
Weights: Qwen/Qwen3.6-35B-A3B- Parameters 36B
- Active 3B · 256 experts, 8 per token
- Native precision BF16
- Layers 40
- KV cache per token 20 KB
- Native context 256K (model card)
- Config positions 256K (config.json max_position_embeddings)
- API context 256K (maker’s API)
- Needs
- ≈ 18.8 GB (weights 17.8 + KV cache 0.2 + overhead 0.9)
- Usable
- 21.6 GB (90% of its memory)
- Generation speed
- 286–429 tok/s (reads the active 3B per token)
- Longest context
- 149K (longer doesn’t fit)
- First token
- about 0.10–0.15 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 67.0 GB | ≈ 70.5 | ○Needs system memory 10–15 tok/s | — |
| Q8_0 | 35.6 GB | ≈ 37.5 | ○Needs system memory 34–51 tok/s | — |
| FP8maker’s file | 34.9 GB | ≈ 36.8 | ○Needs system memory 35–52 tok/s | — |
| Q6_K | 27.5 GB | ≈ 29.0 | ○Needs system memory 54–80 tok/s | — |
| Q5_K_M | 23.8 GB | ≈ 25.2 | ○Needs system memory 93–140 tok/s | — |
| Q5_0 | 23.0 GB | ≈ 24.3 | ○Needs system memory 107–161 tok/s | — |
| Q4_K_M | 20.3 GB | ≈ 21.5 | ●Runs well 253–380 tok/s | 12K |
| Q4_0 | 18.8 GB | ≈ 19.9 | ●Runs well 272–407 tok/s | 93K |
| IQ4_XS | 17.8 GB | ≈ 18.8 | ●Runs well 286–429 tok/s | 149K |
| Q3_K_Mquality drops | 16.4 GB | ≈ 17.3 | ●Runs well 308–463 tok/s | 226K |
| IQ3_XXSquality drops | 12.8 GB | ≈ 13.6 | ●Runs well 383–575 tok/s | 256K |
| Q2_Kquality drops | 12.6 GB | ≈ 13.3 | ●Runs well 390–585 tok/s | 256K |
| IQ2_XXSquality drops | 8.6 GB | ≈ 9.3 | ●Runs well 536–804 tok/s | 256K |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU GeForce RTX 4090; unified memory Apple M5 Max (32-core GPU) · 36 GB.
Run commands
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by us- Epoch Capabilities Index 144
#84 of 118 listed models · CI 142–145.9
- Epoch AI · FrontierMath Tiers 1–3 20.4%
#71 of 84 listed models · reasoning off · ±2.39
Also listed: qwen3.6-35b-a3b 17.5%
- Epoch AI · GPQA Diamond 84.8%
#68 of 168 listed models · reasoning off · ±2.55
Also listed: qwen3.6-35b-a3b 83.8%
- Epoch AI · Mystery Game Puzzles 22%
#41 of 77 listed models · reasoning off · ±4.16
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Oct 28 – Nov 5, 2026
- changes we made
Provider-reported scores
Published by Alibaba Qwen, under the conditions it states| Benchmark | Setting | Score |
|---|---|---|
| SWE-bench Verified | internal agent scaffold, 200K context | 73.4% |
| Terminal-Bench 2.0 | Terminus-2 harness, avg of 5 runs | 51.5% |
| MMLU-Pro | thinking mode | 85.2% |
| GPQA Diamond | thinking mode | 86% |
| Humanity's Last Exam | no tools, thinking mode | 21.4% |
| LiveCodeBench | v6, thinking mode | 80.4% |
| AIME 2026 | full AIME 2026 (I & II), thinking mode | 92.7% |
| MMMU | thinking mode | 81.7% |
Changes
All changes →- 1 price dropped on Alibaba Cloud Model Studio Output (Reasoning on) $1.485
- 2 prices dropped on Alibaba Cloud Model Studio, International Output (Reasoning off) $2.25 · Output (Reasoning on) $2.25
Sources
- Qwen3.6-35B-A3B model card (Hugging Face, Qwen) (opens in a new tab) huggingface.co · checked Sep 26, 2026
- 硅基流动：模型价格in Chinese (opens in a new tab) siliconflow.cn · checked Nov 1, 2026
- 阿里云百炼：模型价格in Chinese (opens in a new tab) help.aliyun.com · checked Nov 2, 2026
- Qwen3.6 35B-A3B model (opens in a new tab) alibabacloud.com · checked Nov 2, 2026
- Alibaba Cloud Model Studio: image and video understanding (opens in a new tab) alibabacloud.com · checked Nov 2, 2026
- Qwen/Qwen3.6-35B-A3B model card (opens in a new tab) huggingface.co · checked Nov 2, 2026
- Qwen/Qwen3.6-35B-A3B model card (opens in a new tab) huggingface.co · checked Nov 2, 2026
- Qwen/Qwen3.6-35B-A3B model card (opens in a new tab) huggingface.co · checked Nov 6, 2026
- Qwen/Qwen3.6-35B-A3B config.json (opens in a new tab) huggingface.co · checked Nov 1, 2026
- Qwen/Qwen3.6-35B-A3B files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3.6-35B-A3B-FP8 files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3.6-35B-A3B README.md (opens in a new tab) huggingface.co · checked Nov 4, 2026
- qwen3.6-35b-a3b - Alibaba Cloud Model Studio (opens in a new tab) docs.modelstudio.console.alibabacloud.com · checked Oct 30, 2026
- Alibaba Cloud Model Studio pricing (opens in a new tab) alibabacloud.com · checked Nov 2, 2026
- 模型上下架与更新-大模型服务平台百炼-阿里云_大模型服务平台百炼(Model Studio)-大模型服务平台百炼(Model Studio)-阿里云帮助中心in Chinese (opens in a new tab) help.aliyun.com · checked Oct 30, 2026
Alibaba Qwen’s official policies: data use, privacy, terms →
Alibaba Qwen · China · qwen.ai
Related models
Snapshots, the same family and other models from this provider