Qwen3.5-397B-A17B
- In Input modalities: TextImageVideo Out Output modalities: Text
- Reasoning
- Reasoning can be off
- Tool use
- Structured output
- Open weights
Prices
5 offerings · prices as listedqwen3.5-397b-a17b| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | $0.172 | $1.032 |
| Prompts over 128K tokens | $0.43 | $2.58 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Nov 2, 2026)
qwen3.5-397b-a17b| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | $0.60 | $3.60 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Nov 2, 2026)
qwen3.5-397b-a17b| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | ¥1.20 | ¥7.20 |
| Prompts over 128K tokens | ¥3.00 | ¥18.00 |
Prices from 阿里云百炼：模型价格 (opens in a new tab) (checked Nov 2, 2026)
qwen3.5-397b-a17b| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 21, 2026 | $0.172 | $1.032 |
| Prompts over 128K tokens no longer listed since Oct 21, 2026 | $0.43 | $2.58 |
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
- Max thinking budget
- 81.92K
- Built-in tools
- Web search
- Licence
- apache-2.0
- Family
- qwen
- Model ID
Not stated on Alibaba Qwen’s pages: reasoning efforts, knowledge cutoff, aliases.
Run locally
Weights: Qwen/Qwen3.5-397B-A17B- Parameters 403.4B
- Active 17B · 512 experts, 10 per token
- Native precision BF16
- Layers 60
- KV cache per token 30 KB
- Native context 256K (model card)
- Config positions 256K (config.json max_position_embeddings)
- API context 256K (maker’s API)
- Needs
- ≈ 209.8 GB (weights 199.6 + KV cache 0.2 + overhead 10.0)
- Usable
- 21.6 GB (90% of its memory)
- Try
- it’s too large for this hardware; see the smallest that runs it below
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 751.4 GB | ≈ 789.2 | ✕Doesn’t fit | — |
| Q8_0 | 399.2 GB | ≈ 419.4 | ✕Doesn’t fit | — |
| FP8maker’s file | 378.3 GB | ≈ 397.4 | ✕Doesn’t fit | — |
| Q6_K | 308.5 GB | ≈ 324.2 | ✕Doesn’t fit | — |
| Q5_K_M | 267.2 GB | ≈ 280.8 | ✕Doesn’t fit | — |
| Q5_0 | 258.3 GB | ≈ 271.4 | ✕Doesn’t fit | — |
| Q4_K_M | 228.2 GB | ≈ 239.9 | ✕Doesn’t fit | — |
| Q4_0 | 211.3 GB | ≈ 222.1 | ✕Doesn’t fit | — |
| INT4 (AWQ / GPTQ)maker’s file | 219.5 GB | ≈ 230.7 | ✕Doesn’t fit | — |
| IQ4_XS | 199.6 GB | ≈ 209.8 | ✕Doesn’t fit | — |
| Q3_K_Mquality drops | 183.6 GB | ≈ 193.0 | ✕Doesn’t fit | — |
| IQ3_XXSquality drops | 143.7 GB | ≈ 151.1 | ✕Doesn’t fit | — |
| Q2_Kquality drops | 140.9 GB | ≈ 148.2 | ✕Doesn’t fit | — |
| IQ2_XXSquality drops | 96.7 GB | ≈ 101.8 | ✕Doesn’t fit | — |
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): no single listed GPU; unified memory Apple M5 Ultra (80-core GPU) · 512 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 147.9
#65 of 118 listed models · CI 145.7–150
- LMArena Text 1,441
#60 of 163 listed models · CI 1,438–1,445 · 97,675 votes
- Epoch AI · FrontierMath Tiers 1–3 31.2%
#65 of 84 listed models · reasoning off · ±2.75
Also listed: qwen3.5-397b-a17b 29.5%
- Epoch AI · GPQA Diamond 86.4%
#61 of 168 listed models · reasoning off · ±2.45
Also listed: qwen3.5-397b-a17b 85.9%
- 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
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Oct 19 – 28, 2026
- EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · License · as of Oct 28, 2026
- changes we made
Provider-reported scores
Published by Alibaba Qwen, under the conditions it states| Benchmark | Setting | Score |
|---|---|---|
| MMLU-Pro | thinking mode | 87.8% |
| GPQA Diamond | thinking mode | 88.4% |
| Humanity's Last Exam | no tools, thinking mode | 28.7% |
| with tools (search agent, context folding) | 48.3% | |
| LiveCodeBench | v6, thinking mode | 83.6% |
| τ²-bench | airline domain with Claude Opus 4.5 system-card fixes | 86.7% |
| BrowseComp | simple context folding | 69% |
| discard-all context strategy | 78.6% | |
| SWE-bench Verified | thinking mode | 76.4% |
| Terminal-Bench 2.0 | thinking mode | 52.5% |
| MMMU | thinking mode | 85% |
Changes
All changes →- Context window 256K → 262.14K
- 1 price dropped on Alibaba Cloud Model Studio Output (Reasoning on) $1.032
- 2 prices dropped on Alibaba Cloud Model Studio, International Output (Reasoning off) $3.60 · Output (Reasoning on) $3.60
Sources
- Qwen3.5-397B-A17B model card (Hugging Face, Qwen) (opens in a new tab) huggingface.co · checked Sep 26, 2026
- Alibaba Cloud Model Studio pricing (opens in a new tab) alibabacloud.com · checked Nov 2, 2026
- 阿里云百炼：模型价格in Chinese (opens in a new tab) help.aliyun.com · checked Nov 2, 2026
- Text generation models - Qwen, DeepSeek, GLM | Model Studio - Alibaba Cloud Model Studio - Alibaba Cloud Documentation Center (opens in a new tab) docs.modelstudio.console.alibabacloud.com · checked Nov 1, 2026
- Qwen3.5 397B-A17B model (opens in a new tab) alibabacloud.com · checked Nov 1, 2026
- Alibaba Cloud Model Studio: newly released models (opens in a new tab) alibabacloud.com · checked Nov 2, 2026
- Qwen/Qwen3.5-397B-A17B model card (opens in a new tab) huggingface.co · checked Nov 2, 2026
- Alibaba Cloud Model Studio model pricing - Alibaba Cloud Model Studio - Alibaba Cloud Documentation Center (opens in a new tab) alibabacloud.com · checked Nov 6, 2026
- Qwen/Qwen3.5-397B-A17B model card (opens in a new tab) huggingface.co · checked Nov 2, 2026
- qwen3.5-397b-a17b (opens in a new tab) docs.modelstudio.console.alibabacloud.com · checked Oct 31, 2026
- Qwen/Qwen3.5-397B-A17B model card (opens in a new tab) huggingface.co · checked Nov 6, 2026
- Qwen/Qwen3.5-397B-A17B config.json (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3.5-397B-A17B files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3.5-397B-A17B-FP8 files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3.5-397B-A17B-GPTQ-Int4 files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3.5-397B-A17B README.md (opens in a new tab) huggingface.co · checked Nov 4, 2026
- Qwen3.5 397B-A17B model (opens in a new tab) alibabacloud.com · checked Nov 1, 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