Qwen3-Next-80B-A3B-Thinking
- In Input modalities: Text Out Output modalities: Text
- Reasoning
- Tool use
- Open weights
Prices
3 offerings · prices as listedqwen3-next-80b-a3b-thinking| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | $0.15 | $1.20 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Oct 11, 2026)
qwen3-next-80b-a3b-thinking| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | $0.144 | $1.434 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Oct 11, 2026)
qwen3-next-80b-a3b-thinking| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | ¥1.00 | ¥10.00 |
Prices from 阿里云百炼：模型价格 (opens in a new tab) (checked Oct 4, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- Structured output
- —
- Knowledge cutoff
- Apr 2025
- Family
- qwen-next-thinking
- Model ID
Not stated on Alibaba Qwen’s pages: max input, reasoning efforts, reasoning can be off, max thinking budget, built-in tools, licence, aliases.
Run locally
Weights: Qwen/Qwen3-Next-80B-A3B-Thinking- Parameters 81.3B
- Active 3B · 512 experts, 10 per token
- Native precision BF16
- Layers 48
- KV cache per token 24 KB
- Native context 256K (model card)
- Config positions 256K (config.json max_position_embeddings)
- API context 128K (maker’s API)
- Needs
- ≈ 64.3 GB (weights 61.0 (maker’s file) + KV cache 0.2 + overhead 3.1)
- Usable
- 21.6 GB (90% of its memory)
- In system memory
- 43.2 GB · 14/48 layers on the GPU (--n-gpu-layers 14)
- Generation speed
- 25–38 tok/s (reads the active 3B per token)
- 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 | 151.5 GB | ≈ 159.3 | ✕Doesn’t fit | — |
| Q8_0maker’s file | 79.0 GB | ≈ 83.1 | ✕Doesn’t fit | — |
| FP8maker’s file | 76.4 GB | ≈ 80.4 | ✕Doesn’t fit | — |
| Q6_Kmaker’s file | 61.0 GB | ≈ 64.3 | ○Needs system memory 25–38 tok/s | — |
| Q5_K_Mmaker’s file | 52.8 GB | ≈ 55.6 | ○Needs system memory 38–57 tok/s | — |
| Q5_0maker’s file | 51.2 GB | ≈ 54.0 | ○Needs system memory 39–59 tok/s | — |
| Q4_K_Mmaker’s file | 45.1 GB | ≈ 47.5 | ○Needs system memory 45–67 tok/s | — |
| Q4_0 | 42.6 GB | ≈ 44.9 | ○Needs system memory 47–71 tok/s | — |
| IQ4_XS | 40.2 GB | ≈ 42.4 | ○Needs system memory 50–75 tok/s | — |
| Q3_K_Mquality drops | 37.0 GB | ≈ 39.1 | ○Needs system memory 58–87 tok/s | — |
| IQ3_XXSquality drops | 29.0 GB | ≈ 30.6 | ○Needs system memory 104–157 tok/s | — |
| Q2_Kquality drops | 28.4 GB | ≈ 30.0 | ○Needs system memory 107–160 tok/s | — |
| IQ2_XXSquality drops | 19.5 GB | ≈ 20.7 | ●Runs well 518–776 tok/s | 48K |
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 NVIDIA H100 SXM; unified memory Apple M5 Max (40-core GPU) · 64 GB.
Run commands
llama-server -hf Qwen/Qwen3-Next-80B-A3B-Thinking-GGUF:Q6_K -c 8192 --n-gpu-layers 14
ollama run hf.co/Qwen/Qwen3-Next-80B-A3B-Thinking-GGUF:Q6_K
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- LMArena Text 1,369
#119 of 163 listed models · CI 1,363–1,375 · 13,476 votes
- LMArena Text · Chinese 1,402
#116 of 161 listed models · CI 1,384–1,420 · 1,059 votes
- LMArena Text · Coding 1,421
#119 of 163 listed models · CI 1,410–1,432 · 2,666 votes
- LMArena Text · Creative writing 1,326
#120 of 163 listed models · CI 1,311–1,340 · 1,776 votes
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Oct 28, 2026
- changes we made
Changes
All changes →- Context window 256K → 131.07K
Sources
- Alibaba Cloud Model Studio pricing (opens in a new tab) alibabacloud.com · checked Oct 11, 2026
- 阿里云百炼：模型价格in Chinese (opens in a new tab) help.aliyun.com · checked Oct 4, 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 Oct 10, 2026
- Qwen3-Next 80B-A3B (Thinking) model (opens in a new tab) alibabacloud.com · checked Oct 8, 2026
- Alibaba Cloud Model Studio: newly released models (opens in a new tab) alibabacloud.com · checked Nov 2, 2026
- Qwen/Qwen3-Next-80B-A3B-Thinking model card (opens in a new tab) huggingface.co · checked Nov 1, 2026
- models.dev: qwen3-next-80b-a3b-thinking (community data, MIT) (opens in a new tab) github.com · checked Sep 30, 2026
- Qwen/Qwen3-Next-80B-A3B-Thinking model card (opens in a new tab) huggingface.co · checked Nov 1, 2026
- Qwen/Qwen3-Next-80B-A3B-Thinking model card (opens in a new tab) huggingface.co · checked Oct 30, 2026
- Qwen/Qwen3-Next-80B-A3B-Thinking config.json (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-Next-80B-A3B-Thinking files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-Next-80B-A3B-Thinking-FP8 files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-Next-80B-A3B-Thinking-GGUF files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-Next-80B-A3B-Thinking README.md (opens in a new tab) huggingface.co · checked Nov 4, 2026
Alibaba Qwen’s official policies: data use, privacy, terms →
Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data
Alibaba Qwen · China · qwen.ai
Related models
Snapshots, the same family and other models from this provider