Hugging FaceFor long documents and questions in many languages, and it can read photos and screenshots too.
- Rating
- 4.2
- Downloads
- 208K
- Parameters
- 122B, 10B active
- Context
- 256K
- Download
- 35 to 245 GB
- Commercial use
- Yes
Too big for this Mac
Its smallest size needs 62 GB of memory, more than M4 · 16 GB can give it.
About
Built so only part of the model runs for each word, with the aim of making it fast and cheap to run. It can think before it answers and can use tools inside Outmake.
How it compares
Outmake estimates it is the 3rd fastest of the 7 chat models it recommends for a 96 GB Mac, and Qwen3.6 35B is about 1.6 times as fast.
Of the chat models Outmake recommends for a 96 GB Mac, LM Arena's public vote rates Gemma 4 31B, Gemma 4 26B and Qwen3.8 27B above it.
On Hugging Face, it was the 6th most downloaded last month of the 7 chat models Outmake recommends for a 96 GB Mac.
Good to know
Its training included practice in million-agent environments with tasks that grew harder step by step.
Its inner layers hold 256 expert sections, and each step uses 8 of them plus 1 shared one.
Sources: Qwen/Qwen3.5-122B-A10B model card, Outmake's speed estimates, LM Arena and Hugging Face. Every line comes from one of these.
Ratings
4.2 out of 5, from how people voted on its answers.
LM Arena ratings of October 28, 2026 (CC BY 4.0): people vote between two answers without knowing which model wrote which. The bar ends at 1488, the best any open model has. The Arena tests the full model; a small size on a Mac answers a little worse.
Good at
Licence
Apache 2.0. Commercial use: yes. The maker's page has the full terms.
Qwen3.5 122B runs well from a Mac with an M1 Max and 64 GB of memory, at UD-IQ2_M, writing about 21 words (28 tokens) a second. 24 of 59 Mac configurations run it well.
Compare with
The nearest models in size, judged for the same Mac.
- GPT-OSS 120BOpenAI's open-weight reasoning model, designed for reasoning problems and agent tasks.Too big for this Mac · 63 GBGet
- Qwen3.8 Flash NextPreview of Qwen's next designToo big for this Mac · 91 GBGet
- MiniMax M2.7Help finding bugs in codeToo big for this Mac · 102 GBGet
- DeepSeek V4 FlashLong documents and tool useToo big for this Mac · 128 GBGet
Sizes
| Size | Download | Memory at 8K | Memory at 32K | Open |
|---|---|---|---|---|
| UD-Q3_K_XLSmaller | 58 GB | 58 GB | 58 GB | Open |
| UD-Q4_K_XLBalanced | 78 GB | 78 GB | 78 GB | Open |
| UD-Q6_K_XLBest | 113 GB | 113 GB | 114 GB | Open |
All 23 sizes
| Size | Download | Memory at 8K | Memory at 32K | Open |
|---|---|---|---|---|
| UD-IQ1_M | 35 GB | 35 GBestimated | 35 GB | Open |
| UD-IQ2_XXS | 38 GB | 38 GBestimated | 38 GB | Open |
| UD-IQ2_M | 40 GB | 40 GBestimated | 40 GB | Open |
| UD-Q2_K_XL | 43 GB | 43 GBestimated | 43 GB | Open |
| UD-IQ3_XXS | 46 GB | 46 GBestimated | 46 GB | Open |
| UD-IQ3_S | 47 GB | 47 GBestimated | 48 GB | Open |
| Q3_K_S | 53 GB | 53 GBestimated | 54 GB | Open |
| Q3_K_M | 57 GB | 57 GBestimated | 58 GB | Open |
| UD-IQ4_XS | 61 GB | 61 GBestimated | 61 GB | Open |
| UD-IQ4_NL | 62 GB | 62 GBestimated | 63 GB | Open |
| Q4_K_S | 73 GB | 73 GBestimated | 73 GB | Open |
| MXFP4_MOE | 76 GB | 76 GBestimated | 76 GB | Open |
| Q4_K_M | 77 GB | 77 GBestimated | 78 GB | Open |
| Q5_K_S | 87 GB | 87 GBestimated | 88 GB | Open |
| Q5_K_M | 92 GB | 92 GBestimated | 93 GB | Open |
| UD-Q5_K_XL | 93 GB | 93 GBestimated | 93 GB | Open |
| Q6_K | 102 GB | 102 GBestimated | 102 GB | Open |
| Q8_0 | 131 GB | 131 GBestimated | 131 GB | Open |
| UD-Q8_K_XL | 172 GB | 172 GBestimated | 172 GB | Open |
| BF16 | 245 GB | 245 GBestimated | 246 GB | Open |
Memory is what llama.cpp's own calculator reported for each size when the catalog was built, with the app's settings. Smaller, Balanced and Best are the three sizes the store names.
Run It Anywhere
Qwen3.5 122B is an open model, so any app that runs GGUF files can run it. These take the UD-Q4_K_XL size from unsloth straight from Hugging Face.
- LM StudioMac, Windows, Linux
Open - JanMac, Windows, Linux
Open - MstyMac, Windows, Linux
Open - OllamaMac, Windows, Linux
ollama run hf.co/unsloth/Qwen3.5-122B-A10B-GGUF:UD-Q4_K_XL - llama.cppMac, Windows, Linux
llama-server -hf unsloth/Qwen3.5-122B-A10B-GGUF:UD-Q4_K_XL - Docker Model RunnerMac, Windows, Linux
docker model run hf.co/unsloth/Qwen3.5-122B-A10B-GGUF:UD-Q4_K_XL
On a PC Graphics Card
| Video memory | Largest size that fits | Download | Memory at 8K |
|---|---|---|---|
| 6 GB | None worth running | needs 40 GB | |
| 8 GB | None worth running | needs 40 GB | |
| 10 GB | None worth running | needs 40 GB | |
| 11 GB | None worth running | needs 40 GB | |
| 12 GB | None worth running | needs 40 GB | |
| 16 GB | None worth running | needs 40 GB | |
| 20 GB | None worth running | needs 40 GB | |
| 24 GB | None worth running | needs 40 GB | |
| 32 GB | None worth running | needs 40 GB | |
| 48 GB | UD-IQ3_XXS | 46 GB | 46 GB |
The largest size, of those that keep three bits a weight or more, whose memory at an 8K chat is under 92% of the card's; the rest is the desktop's. Estimated from the files, the same way as for a Mac; speed on a PC is not estimated yet. A size that doesn't fit the card still runs, slower, with part of it in ordinary memory.
Tests
MMLU-Pro: graduate-level questions in 14 subjects. GPQA Diamond: questions PhD scientists write for each other. Humanity's Last Exam: questions written to be beyond today's models. SWE-bench Verified: real GitHub issues fixed so the project's tests pass. Terminal-Bench 2.0: jobs finished alone in a terminal. Scores are the ones filed on the model's Hugging Face page or the board's own table, read 2026-11-05; some come from the maker.
On every Mac
| Mac | Size | Fit | Memory | Writes | Reads |
|---|---|---|---|---|---|
| M6 · 16 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M6 · 24 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M6 · 32 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M5 · 16 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M5 · 24 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M5 · 32 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M5 Pro · 24 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M5 Pro · 48 GB | UD-IQ2_XXS | Tight fit | 38 GB | 28 | 325 |
| M5 Pro · 64 GB | UD-IQ2_M | Runs well | 42 GB | 27 | 406 |
| M5 Max · 36 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M5 Max · 48 GB | UD-IQ2_XXS | Tight fit | 38 GB | 50 | 813 |
| M5 Max · 64 GB | UD-IQ2_M | Runs well | 42 GB | 48 | 813 |
| M5 Max · 128 GB | UD-Q3_K_XL | Runs great | 62 GB | 37 | 813 |
| M5 Ultra · 96 GB | UD-Q3_K_XL | Runs well | 62 GB | 65 | 1300 |
| M5 Ultra · 256 GB | UD-Q6_K_XL | Runs great | 117 GB | 40 | 1625 |
| M5 Ultra · 512 GB | UD-Q6_K_XL | Runs great | 117 GB | 40 | 1625 |
| M4 · 16 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M4 · 24 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M4 · 32 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M4 Pro · 24 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M4 Pro · 48 GB | UD-IQ2_XXS | Tight fit | 38 GB | 20 | 109 |
| M4 Pro · 64 GB | UD-IQ2_M | Runs well | 42 GB | 19 | 109 |
| M4 Max · 36 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M4 Max · 48 GB | UD-IQ2_XXS | Tight fit | 38 GB | 30 | 273 |
| M4 Max · 64 GB | UD-IQ2_M | Runs well | 42 GB | 29 | 273 |
| M4 Max · 96 GB | UD-Q3_K_XL | Runs well | 62 GB | 20 | 218 |
| M4 Max · 128 GB | UD-Q3_K_XL | Runs great | 62 GB | 25 | 273 |
| M3 · 8 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M3 · 16 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M3 · 24 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M3 Pro · 18 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M3 Pro · 36 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M3 Max · 36 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M3 Max · 48 GB | UD-IQ2_XXS | Tight fit | 38 GB | 27 | 226 |
| M3 Max · 64 GB | UD-IQ2_M | Runs well | 42 GB | 26 | 226 |
| M3 Max · 96 GB | UD-Q3_K_XL | Runs well | 62 GB | 17 | 170 |
| M3 Max · 128 GB | UD-Q3_K_XL | Runs great | 62 GB | 21 | 226 |
| M3 Ultra · 96 GB | UD-Q3_K_XL | Runs well | 62 GB | 33 | 339 |
| M3 Ultra · 256 GB | UD-Q6_K_XL | Runs great | 117 GB | 23 | 339 |
| M3 Ultra · 512 GB | UD-Q6_K_XL | Runs great | 117 GB | 23 | 452 |
| M2 · 8 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M2 · 16 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M2 · 24 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M2 Pro · 16 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M2 Pro · 32 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M2 Max · 32 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M2 Max · 64 GB | UD-IQ2_M | Runs well | 42 GB | 26 | 166 |
| M2 Max · 96 GB | UD-Q3_K_XL | Runs well | 62 GB | 21 | 211 |
| M2 Ultra · 64 GB | UD-IQ2_M | Runs well | 42 GB | 38 | 333 |
| M2 Ultra · 128 GB | UD-Q3_K_XL | Runs great | 62 GB | 33 | 333 |
| M2 Ultra · 192 GB | UD-Q6_K_XL | Runs great | 117 GB | 23 | 333 |
| M1 · 8 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M1 · 16 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M1 Pro · 16 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M1 Pro · 32 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M1 Max · 32 GB | UD-Q3_K_XL | Too big | 62 GB | ||
| M1 Max · 64 GB | UD-IQ2_M | Runs well | 42 GB | 21 | 123 |
| M1 Ultra · 64 GB | UD-IQ2_M | Runs well | 42 GB | 28 | 247 |
| M1 Ultra · 128 GB | UD-Q3_K_XL | Runs great | 62 GB | 25 | 247 |
Writes and reads are words a second (a token is about three quarters of a word), 4K tokens into a chat, estimated unless marked measured. The size is the one the store offers that Mac. Constants fitted October 11, 2026; how, and how wrong.
Badge for the model card
Shows on Qwen3.5 122B's Hugging Face card which Macs run it well, linked to this page. It updates every morning.
Sizes and files are read from unsloth/Qwen3.5-122B-A10B-GGUF on Hugging Face and checked every morning. The rating is the model's score on the LM Arena text leaderboard (CC BY 4.0), where people vote between two answers without knowing which model wrote which. Catalog of November 4, 2026.