whatmodel.sh / Qwen3-Coder-Next (80B-A3B)
Qwen3-Coder-Next (80B-A3B)
↗ share
+ contribute
Qwen3-Coder-Next (80B-A3B) is an open-weights language model from Alibaba, released February 2026. Its published size is 79.7B parameters (3B active per token) and a 262k-token context window. Licence: apache-2.0.
$ how it compares
The LLMs ranked either side of Qwen3-Coder-Next (80B-A3B), out of 828 with a score, then more from Alibaba.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 133 |  Gemma 4 31B | 67±1 | api $0.19/M +local | ||
| 134 |  Gemini 3.1 Flash-Lite | 67±1 | api $0.56/M | ||
| 135 | Qwen3-Coder-Next (80B-A3B) | 67±2 | api $0.60/M +local | ||
| 136 | Gemini 2.5 Pro Preview (03-25) | 67±2 | comparison only | ||
| 137 | Gemini 2.5 Pro Preview (05-06) | 67±1 | comparison only | ||
| more from Alibaba | |||||
| 18 | Qwen3.8 Max (2.4T-A95B) | 78±1 | api $3.00/M +local | ||
| 34 | Qwen3.8 Max (0902) | 76±1 | api $3.00/M | ||
| 37 | Qwen3.8-Flash-Next | 75±1 | local | ||
$ what it's good at
agent tasksnot ranked on this task yet · 4 tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| SWE-bench Verified lab | 70.6% ✎ | ▁▁▁▁▁▁▂▄█▆▁▁ beats 79/131 | — |
| SWE-bench Multilingual lab | 62.8% ✎ | ▁··▁▃▂▅▇█▃▃▁ beats 22/60 | — |
| SWE-bench Pro (split unspecified) lab | 44.3% ✎ | ▃·▁▁▁▂▂▃█▇▃▁ beats 21/55 | — |
| Terminal-Bench 2.0 lab | 36.2% ✎ | ▂▂▅▅████▆▂▃▃ beats 10/36 | — |
coding▁▁▁▃▆██▇▆▄▃▁ #173 of 458 * · 8.6% of its score · 4 tests
Trails QwQ-32B-Preview, Ministral 3 8B Reasoning 2512, Phi-4-reasoning 14B here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| HumanEval (Vector Institute, Inspect Evals) | 91.5% ✎ | ▄·▄·█·█·█·▄▄ beats 5/9 | 8.6% |
| SWE-bench Verified lab | 70.6% ✎ | ▁▁▁▁▁▁▂▄█▆▁▁ beats 79/131 | — |
| SWE-bench Multilingual lab | 62.8% ✎ | ▁··▁▃▂▅▇█▃▃▁ beats 22/60 | — |
| SWE-bench Pro (split unspecified) lab | 44.3% ✎ | ▃·▁▁▁▂▂▃█▇▃▁ beats 21/55 | — |
math▁▂▃▅▆▅▄█▄▁▁▁ #63 of 398 * · 17% of its score · 1 test
Beats Qwen3.6 Max Preview, Claude Sonnet 4.6, Claude Opus 4.5 here, though they rank higher overall. Trails INTELLECT-3 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| MATH (Vector Institute, Inspect Evals) | 97.8% ✎ | ▄··██▄··▄·▄█ beats 9/9 | 17% |
other tests3 not in a task · 73% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond (Vector Institute, Inspect Evals) | 75.6% ✎ | ▃··▃▃▃▃·▃·▃█ beats 9/9 | 39% |
| DSEffi-Bench Pass (1,000 data-science tasks, greedy) | 50.5% ✎ | ▅··▃▅·█·▃██▃ beats 8/15 | 19% |
| MMLU-Pro (Vector Institute, Inspect Evals) | 82.5% ✎ | ▃····▃▅·▃▅·█ beats 7/9 | 16% |
$ where to run it
Through an API
api · Alibaba Cloud Model Studio · $0.60/M · $0.3 in / $1.5 out per 1M tokens · 262k context · www.alibabacloud.com ↗ · price sourceUSD per 1M tokens, Alibaba Cloud Model Studio, Singapore (International deployment scope), qwen3-coder-next, lowest input tier (0 < input tokens per request ≤ 32K). Requests with 32K–128K input are billed $0.5/$2.5 and 128K–256K $0.8/$4. Free quota and context-cache discounts excluded. Currency: USD. Alibaba Cloud official Model Studio pricing page, observed 2026-09-25.
llama.cpp · recommended
- Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.macOS / Linux
curl -LsSf https://llama.app/install.sh | shWindowsirm https://llama.app/install.ps1 | iex - Run Qwen3-Coder-Next (80B-A3B) · needs ~24.41 GB · from huggingface.co/unslothchat in the terminal
llama cli -hf unsloth/Qwen3-Coder-Next-GGUF:UD-Q2_K_XL - Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf unsloth/Qwen3-Coder-Next-GGUF:UD-Q2_K_XL
LM Studio
version
- Install LM Studio · Open the app once before using the lms command.
- Run Qwen3-Coder-Next (80B-A3B) · needs ~44.8 GB, Mac only · from huggingface.co/lmstudio-community
- Or download it in the terminal
lms get https://huggingface.co/lmstudio-community/Qwen3-Coder-Next-MLX-4bit -y - Then run lms ls to see its name, and lms chat followed by that name
Ollama
- Install Ollama
download it (macOS / Windows) ↗
Linuxcurl -fsSL https://ollama.com/install.sh | sh - Run Qwen3-Coder-Next (80B-A3B) · needs ~48.32 GB · from ollama.comchat in the terminal
ollama run qwen3-coder-next
MLX (Mac)
version
- Install MLX (Mac) · Apple silicon Macs only, text only.with uv
uv tool install mlx-lmorpip install mlx-lm - Run Qwen3-Coder-Next (80B-A3B) · needs ~44.8 GB, Mac only · from huggingface.co/lmstudio-communitychat in the terminal
mlx_lm.chat --model lmstudio-community/Qwen3-Coder-Next-MLX-4bit - Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model lmstudio-community/Qwen3-Coder-Next-MLX-4bit
Official weights and documentation: huggingface.co/Qwen ↗
$ how the score was worked out · Vector Institute Evaluation L… 80%, DSEffi-Bench: Demystifying La… 19%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 Vector Institute Evaluation Leaderboard (Inspect Evals runs) | 4 | 80% |
 DSEffi-Bench: Demystifying Large Language Models' Capability in Efficient Data Science Code Generation — authors' own model runs | 1 | 19% |
| prior · pull toward the middle | 1% | |
| weighted average | = 66.9 |
not counted (4)
| Qwen/Qwen3-Coder-Next model card | SWE-bench Verified [correct] | 70.6% | fewer than three models have this test yet, or this source ran it for this model alone |
| Qwen/Qwen3-Coder-Next model card | SWE-bench Multilingual [correct] | 62.8% | fewer than three models have this test yet, or this source ran it for this model alone |
| Qwen/Qwen3-Coder-Next model card | SWE-bench Pro (split unspecified) [correct] | 44.3% | fewer than three models have this test yet, or this source ran it for this model alone |
| Qwen/Qwen3-Coder-Next model card | Terminal-Bench 2.0 [correct] | 36.2% | fewer than three models have this test yet, or this source ran it for this model alone |
$ published speed measurements · 1
Reference results for the hardware, engine and settings below. They aren't verified speeds of the downloads above or predictions for your machine.
| AMD Ryzen AI Max+ 395 (Strix Halo), Radeon 8060S iGPU; Beelink GTR9 Pro 128GB | IQ4_XS | 61.91 tok/s | llama.cpp 1fd5f4803 (build_number 134 as recorded), Vulkan (RADV) · llama-bench; tg128; 128 generated tokens; context depth 0; n_prompt=0 row; n_batch=2048; n_ubatch=512; flash_attn=1; type_k=f16; type_v=f16; n_gpu_layers=999; threads=16; one sequence; file label r10 repetitions Author's own llama-bench CSV row (avg_ts mean over repetitions, stddev_ts retained). GPU string as recorded: Radeon 8060S Graphics (RADV STRIX_HALO). Unified memory; the guide reports ~124GiB OS-visible. Kernel/Mesa/power profile per the run folder notes. Not prompt processing or multi-request throughput. Curated in the guide's data/benchmarks.csv line 44 (2026-06-02, measured-local): Modern Qwen coding-model row repeated on b9467 after Reddit feedback; useful for current-model capability context, but it does not replace the older Qwen3-Coder 30B Q4_K_S speed-first headline. | source ↗ |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json