[whatmodel.sh](../../README.md) / Qwen3-Next 80B-A3B Thinking
# Qwen3-Next 80B-A3B Thinking
Alibaba · released 2025-09 · 80B params (3B active) · apache-2.0
**score 62.7** ±0.6 · #174 of 732 · from 52 benchmark results across 9 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Qwen3-Next 80B-A3B Thinking among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆█●▅▅▄▄▂▁ #158 of 313 15% 8
coding ▁▁▁▃▅▇█●▅▄▃▁ #147 of 469* 3.0% 1
chat & writing ▁▁▁▃▆▄▆█●▇▇▃ #143 of 402 8.8% 3
math ▁▁▃▅█▆▄●▄▂▁▁ #130 of 431 12% 5
agent tasks · #158 of 313 · 15% of its score
Beats [o1 (2024-12-17)](o1-2024-12-17.md), [o4-mini](o4-mini.md), [o3-mini](o3-mini.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Jamba 1.5 Mini](jamba-1-5-mini.md), [Jamba 1.5 Large](jamba-1-5-large.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
τ²-bench Telecom (maker) 43.9% ▁·▅█●▅▂▃▃▂▃▁ beats 37/89 very hard · avg 9.2% 5.3% sharp · at its level
τ-bench Retail (maker) 69.6% ▁····▃·▂▇●▂▆ beats 20/28 medium · avg 37% 2.4% at its level
τ²-bench Retail (maker) 67.8% ▁····▂▁▃▆●▃▁ beats 23/44 medium · avg 37% 2.2% at its level
τ²-bench Airline (maker) 60.5% ▁▁·▂▁▃▂▁█●▂▃ beats 33/42 hard · avg 32% 1.6% at its level
τ-bench Airline (maker) 49.0% ▁▁▁·▅▄▁█●▁▁▇ beats 20/28 hard · avg 22% 1.6% at its level
BFCL V4 (December 2025 archive) (… 49.7% ▁▁▁▂▆▆▅▆●▂▄▃ beats 71/114 medium · avg 35% 0.9% at its level
τ²-bench aggregate (maker) 57.4% ▁▁▁·▁▁●▃▃▄█▁ beats 8/39 medium · avg 40% 0.7% at its level · noisy
BFCL v3 (maker) 72.0% ▁▁▁▁▁▃▅█▆▅●▁ beats 160/171 medium · avg 59% 0.6% blunt · at its level
coding · #147 of 469 · 3.0% of its score
Beats [Qwen3-Coder-Next (80B-A3B)](qwen3-coder-next.md), [Gemini 2.5 Pro](gemini-2-5-pro.md), [Gemini 3.1 Flash-Lite](gemini-3-1-flash-lite.md) here, though they rank higher overall. Trails [Llama 3.1 Nemotron Ultra 253B v1](llama-3-1-nemotron-ultra-253b.md), [Devstral Small 2 24B](devstral-small-2.md), [Gemma 4 12B](gemma-4-12b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LiveCodeBench v6 (Feb–May 2025) (… 68.7% ▁▁▃▁█▅▄●▅▅▃█ beats 19/38 medium · avg 41% 3.0% at its level · several runs
chat & writing · #143 of 402 · 8.8% of its score
Beats [o3-mini](o3-mini.md), [INTELLECT-3](intellect-3.md), [MiniMax-M2](minimax-m2.md) here, though they rank higher overall. Trails [LFM2.5 8B-A1B](lfm2-5-8b-a1b.md), [LFM2.5 2.6B](lfm2-5-2-6b.md), [Spark-X2.5 1.7B](spark-x2-5-1-7b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LMArena (text) 1369.2 ▁▁▁▂▃▃▃▆●▇█▃ beats 212/385 avg 1260 4.1% blunt
Arena-Hard v2 aggregate (maker) 62.3% ▁▁··▄▃▁▃●▁▁▇ beats 11/22 hard · avg 28% 3.7% sharp · at its level
IFEval (aggregation unspecified) … 88.9% ▁▂▄▂▂▄▆█●▅▂▁ beats 103/145 easy · avg 78% 1.0% typical
math · #130 of 431 · 12% of its score
Beats [Claude Sonnet 4.6](claude-sonnet-4-6.md), [MiniMax-M3](minimax-m3.md), [MiMo-V2.5 (310B-A15B)](mimo-v2-5.md) here, though they rank higher overall. Trails [Olmo 3.1 32B Think](olmo-3-1-32b-think.md), [Qwen3-VL 30B-A3B Thinking](qwen3-vl-30b-a3b-thinking.md), [Granite 4.2 8B](granite-4-2-8b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
HMMT 2025 (edition unspecified) (… 73.9% ▃█▇▅●▅▁·▁··▁ beats 20/32 hard · avg 34% 3.9% sharp · at its level
PolyMATH (maker) 56.3% ▁···▃▁·▇█▄●▄ beats 18/23 hard · avg 32% 2.9% at its level · several runs
AIME 2025 (maker) 87.8% ▁·▁▁▁▂▅▆█●▅▁ beats 106/133 medium · avg 51% 2.3% sharp · small run
HMMT February 2025 (maker) 73.7% ▁▁▁▃▁▅●▃██▁▄ beats 11/31 medium · avg 49% 1.8% typical
HMMT November 2025 (maker) 81.2% ▁▁····▁●▃█▇▁ beats 6/23 medium · avg 62% 1.5% typical
other tests (30)
test result vs the others in its run difficulty counts why
Medmarks-V mean win rate (19 veri… 0.5888 ▁▄▃▃▇▄█▃▅●▁▁ beats 28/30 avg 0.418 7.6% sharp
GPQA Diamond (maker) 77.2% ▁▁▁▁▂▄▅▅█●█▄ beats 191/322 hard · avg 47% 5.1% sharp · at its level
GPQA (subset unspecified) (maker) 77.2% ▂▁▂▂▄▃▆▃▅█●▃ beats 92/106 hard · avg 49% 4.2% sharp · at its level
INCLUDE (maker) 78.9% ▂····▂▂▄▄▄●▄ beats 24/36 medium · avg 59% 3.8% at its level · several runs
MMLU-Pro (maker) 82.7% ▁▁·▁▁▁▃▂▄▅●▁ beats 152/188 medium · avg 61% 3.7% several runs
Medmarks-V weighted mean win rate… 58.9% ▂▃▅▂▆▃█▅█▅●▆ beats 29/34 medium · avg 41% 3.3% blunt · at its level
KINA (Knowledge Index of Noah's A… 28.3% ▁▁▂▁▃▄●▆▄▅▂▃ beats 15/39 very hard · avg 21% 2.7% typical
BRIDGE clinical text, zero-shot a… 43.9% ▁▁▁▁▁▂▂▃▄▆█● beats 90/97 hard · avg 31% 2.6% blunt · at its level
MMLU-ProX (maker) 78.7% ▁·▁▁▁▂▁▃▃●▅▄ beats 20/36 medium · avg 61% 2.6% several runs
LongBench v2 (maker) 48.0% ▂·▁·▂▃▁▄●▃█▄ beats 16/33 very hard · avg 35% 2.5% at its level
C-Eval (maker) 89.7% ▁···▁▁▂▁▂▂▃● beats 21/33 medium · avg 73% 2.2% typical
MMMLU (maker) 81.3% ▂▁▁▂▃▂▄▄▅●█▃ beats 32/46 medium · avg 63% 2.0% at its level
MMLU-Redux (maker) 92.5% ▁▁▃▁▂▃▄▅▇▅●▄ beats 47/63 easy · avg 82% 1.9% typical
OJBench (maker) 29.7% ▁···▁▃·▅▄▅●█ beats 16/26 very hard · avg 13% 1.9% at its level
Global PIQA (maker) 83.5% ▃▃··▃··▃▃●█▅ beats 6/11 medium · avg 71% 1.5% at its level
VitaBench (maker) 29.5% ▄··▂▄▆▆●▄▄▄▂ beats 13/21 very hard · avg 11% 1.5% at its level
MultiChallenge (maker) 51.3% ▁·▁·▁▃▆▅●▂█▅ beats 15/30 hard · avg 31% 1.4% at its level
IFBench (aggregation unspecified)… 61.5% ▂▆▂▄▄▄▂●▄▆█▇ beats 45/86 medium · avg 39% 1.4% at its level
SuperGPQA (maker) 60.8% ▂▁▂▂▅▆▃▇▇●█▃ beats 36/46 medium · avg 39% 1.4% at its level
MAXIFE (maker) 79.9% ▂·▂·▂···●·▄█ beats 5/11 medium · avg 64% 1.3% typical
WritingBench (maker) 84.6% ▁·▁▂▁··▅●▂▂▁ beats 9/20 easy · avg 71% 1.2% typical
LiveBench 2024-11-25 (maker) 76.6% ▄··▂·▂·▂█●▂▂ beats 9/13 medium · avg 61% 1.1% typical
Multi-IF (aggregation unspecified… 77.8% ▁▁▁▁▁▁▃▄●▄▁▂ beats 27/41 medium · avg 63% 1.0% typical
NOVA-63 (maker) 53.3% ▃▃·▃···▃●▃▃█ beats 4/11 medium · avg 39% 0.9% blunt · at its level
EuroEval HellaSwag (English), tes… 88.1% ▃·▁▁▂▁▂▅█▇●▁ beats 87/100 medium · avg 67% 0.4% sharp · noisy
DeepPlanning (maker) 0.4% ●·····▂··▄█▄ beats 0/9 very hard · avg 7.1% 0.4% noisy
EuroEval Life in the UK (English … 95.9% ▁▁▁▁▁▁▁▂▄█▄● beats 145/146 easy · avg 82% 0.2% noisy
WMT24++ (maker) 57.4% ▁·▁▁·●·▁▂▆█▃ beats 3/27 easy · avg 66% 0.1% blunt · at its level
Vectara hallucination leaderboard… 96.3% ▁▁▂▃▄▄█●▃▂▂▁ beats 94/140 very easy · avg 94% 0.1% blunt · maxed out · noisy
Vectara hallucination leaderboard… 90.7% ▁▂▁▂▅●▂▃▄▃▁▁ beats 52/98 very easy · avg 90% 0.0% blunt
source results share of the score
Qwen/Qwen3-Next-80B-A3B-Thinking model … 22 45% ▇▇▇▇▇
Qwen/Qwen3.5-9B model card (submitted /… 18 24% ▇▇▇
Medmarks: A Comprehensive Open-Source L… 2 11% ▇
Qwen/Qwen3.5-4B model card (submitted /… 3 9% ▇
LMArena leaderboard dataset 1 4% ▇
Knowledge Index of Noah's Ark 1 3% ▇
BRIDGE Medical Leaderboard 1 3% ▇
EuroEval leaderboards 2 1% ▇
Vectara hallucination leaderboard 2 0% ▇
prior: pull toward the middle 0% ▇
weighted average = 62.7
Licences: Qwen/Qwen3-Next-80B-A3B-Thinking model card ([apache-2.0](../../licenses/Qwen--Qwen3-Next-80B-A3B-Thinking.txt)) · Qwen/Qwen3.5-9B model card ([apache-2.0](../../licenses/Qwen--Qwen3.5-9B.txt)) · Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks ([CC BY 4.0](../../licenses/paper-2605.01417v1-CC-BY-4.0.txt)) · Qwen/Qwen3.5-4B model card ([apache-2.0](../../licenses/Qwen--Qwen3.5-4B.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Knowledge Index of Noah's Ark ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · BRIDGE Medical Leaderboard ([MIT](../../licenses/bridge-medical-leaderboard-MIT.txt)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE)) · Vectara hallucination leaderboard ([Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0))
## How to run it
### Through an API
api (Alibaba Cloud Model Studio) · $0.41/M blended · $0.15 in / $1.2 out per 1M tokens · [alibabacloud.com](https://www.alibabacloud.com/help/en/model-studio/models)
[price source](https://www.alibabacloud.com/help/en/model-studio/model-pricing): USD per 1M tokens, Singapore (International), qwen3-next-80b-a3b-thinking, thinking mode only. Observed 2026-09-25. · [context source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/models/alibaba/qwen3-next-80b-a3b-thinking.toml)
### llama.cpp (recommended)
1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md)
```sh
# macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
# Windows
irm https://llama.app/install.ps1 | iex
```
2. Run Qwen3-Next 80B-A3B Thinking (ud-q2_k_xl, needs ~24.5 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Qwen3-Next-80B-A3B-Thinking-GGUF/tree/main))
```sh
llama cli -hf unsloth/Qwen3-Next-80B-A3B-Thinking-GGUF:UD-Q2_K_XL
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf unsloth/Qwen3-Next-80B-A3B-Thinking-GGUF:UD-Q2_K_XL
```
### LM Studio
1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download)
2. Run Qwen3-Next 80B-A3B Thinking (4bit, needs ~44.8 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Qwen3-Next-80B-A3B-Thinking-MLX-4bit/tree/740e760a50c07f96cb3fb62bd1d861382fd0a03a))
```sh
lms get https://huggingface.co/lmstudio-community/Qwen3-Next-80B-A3B-Thinking-MLX-4bit -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: ud-q2_k_xl (~24.5 GB) `lms get https://huggingface.co/unsloth/Qwen3-Next-80B-A3B-Thinking-GGUF@UD-Q2_K_XL -y` · 8bit (~84.7 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Qwen3-Next-80B-A3B-Thinking-MLX-8bit -y`
### Ollama
1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download)
```sh
# Linux
curl -fsSL https://ollama.com/install.sh | sh
```
2. Run Qwen3-Next 80B-A3B Thinking (Q4_K_M, needs ~48.5 GB, from [ollama.com](https://ollama.com/library/qwen3-next/tags))
```sh
ollama run qwen3-next:80b-a3b-thinking-q4_K_M
```
### MLX (Mac)
1. Install MLX (Mac). Apple silicon Macs only, text only.
```sh
# with uv
uv tool install mlx-lm
# or
pip install mlx-lm
```
2. Run Qwen3-Next 80B-A3B Thinking (4bit, needs ~44.8 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Qwen3-Next-80B-A3B-Thinking-MLX-4bit/tree/740e760a50c07f96cb3fb62bd1d861382fd0a03a))
```sh
mlx_lm.chat --model lmstudio-community/Qwen3-Next-80B-A3B-Thinking-MLX-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model lmstudio-community/Qwen3-Next-80B-A3B-Thinking-MLX-4bit
```
Other versions: 8bit (~84.7 GB, Mac only) `mlx_lm.chat --model lmstudio-community/Qwen3-Next-80B-A3B-Thinking-MLX-8bit`
Model repository and documentation: [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Thinking)
## Quantization evidence
How much of the full-precision result a quantized version kept, measured within one study. Rows from the same study share a table and aren't independent. They don't describe every download of that precision.
precision test full → quantized kept
FP8 (Qwen official checkpoint… GPQA Diamond (reasoning table; pe… 0.737 → 0.768 104.1%
FP8 (Qwen official checkpoint… MATH-500 (reasoning table; per si… 0.94 → 0.876 93.2%
FP8 (Qwen official checkpoint… AIME 2025 (reasoning table; per s… 0.6 → 0.5 83.3%
FP8 block (E4M3; weights stat… GPQA Diamond (reasoning table; pe… 0.737 → 0.768 104.1%
FP8 block (E4M3; weights stat… MATH-500 (reasoning table; per si… 0.94 → 0.944 100.4%
FP8 block (E4M3; weights stat… AIME 2025 (reasoning table; per s… 0.6 → 0.5 83.3%
FP8 W8A8 dynamic (E4M3; weigh… GPQA Diamond pass@1 (n=8; lightev… 0.737 → 0.748 101.4%
FP8 W8A8 dynamic (E4M3; weigh… GSM8K 5-shot strict-match (OpenLL… 0.846 → 0.848 100.2%
FP8 W8A8 dynamic (E4M3; weigh… MMLU 5-shot acc (OpenLLM v1 confi… 0.85 → 0.851 100.1%
FP8 W8A8 dynamic (E4M3; weigh… MMLU-Pro 5-shot acc (OpenLLM v2 l… 0.727 → 0.728 100.1%
FP8 W8A8 dynamic (E4M3; weigh… OpenLLM v1 average (6 tasks: ARC-… 0.732 → 0.732 100.0%
FP8 W8A8 dynamic (E4M3; weigh… OpenLLM v2 average (6 tasks: IFEv… 0.391 → 0.384 98.3%
FP8 W8A8 dynamic (E4M3; weigh… IFEval inst-level strict acc 0-sh… 0.448 → 0.424 94.7%
FP8 W8A8 dynamic (E4M3; weigh… Reasoning average (3 tasks: AIME … 0.759 → 0.705 92.9%
FP8 W8A8 dynamic (E4M3; weigh… MATH-500 pass@1 (n=8; lighteval, … 0.94 → 0.868 92.3%
FP8 W8A8 dynamic (E4M3; weigh… AIME 2025 pass@1 (n=8; lighteval,… 0.6 → 0.5 83.3%
INT4 W4A16 (GPTQ, symmetric, … IFEval inst-level strict acc 0-sh… 0.448 → 0.46 102.7%
INT4 W4A16 (GPTQ, symmetric, … GPQA Diamond pass@1 (n=8; lightev… 0.737 → 0.737 100.0%
INT4 W4A16 (GPTQ, symmetric, … OpenLLM v1 average (6 tasks: ARC-… 0.732 → 0.729 99.6%
INT4 W4A16 (GPTQ, symmetric, … MMLU 5-shot acc (OpenLLM v1 confi… 0.85 → 0.846 99.5%
INT4 W4A16 (GPTQ, symmetric, … OpenLLM v2 average (6 tasks: IFEv… 0.391 → 0.389 99.5%
INT4 W4A16 (GPTQ, symmetric, … GSM8K 5-shot strict-match (OpenLL… 0.846 → 0.838 99.0%
INT4 W4A16 (GPTQ, symmetric, … MMLU-Pro 5-shot acc (OpenLLM v2 l… 0.727 → 0.714 98.3%
INT4 W4A16 (GPTQ, symmetric, … MATH-500 pass@1 (n=8; lighteval, … 0.94 → 0.872 92.8%
INT4 W4A16 (GPTQ, symmetric, … Reasoning average (3 tasks: AIME … 0.759 → 0.703 92.6%
INT4 W4A16 (GPTQ, symmetric, … AIME 2025 pass@1 (n=8; lighteval,… 0.6 → 0.5 83.3%
Each kept value links to its source.
[interactive version](https://whatmodel.sh/model/qwen3-next-80b-a3b-thinking) · [JSON](https://whatmodel.sh/api?format=json&all=1)