LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
Qwen3.5 122B A10B
◆ OPEN
by Alibaba
OPEN WEIGHTSLOCALAPILONG CONTEXTPAIDREASONING
RANKED #78 OF 316 ASSISTANTS · OVERALL #105 OF 493 RANKED · VIBE SCORE 6.1 · 157 VOICES ·
COPY BADGE pastes a README snippet with the rank SVG.
The Qwen3.5 122B-A10B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency. In terms of...
HAPPY9.9
21 mentions
SAD5.1
6 mentions
VIBE TRAJECTORY
26 weeks · 148 voices in this window
HONEYMOON FADING
Early praise is cooling off in recent weeks.
HOW TO USE IT
Open weights. You can use a hosted service, or download it and run it yourself, free.
RUN IT ON YOUR OWN MACHINE
Open weights: download Qwen3.5 122B A10B once and it's yours. No subscription, no rate limits, works offline.
122B total parameters (10B active MoE), a Q4_K_M quant file is about 72 GB
- BARE MINIMUM
Apple Mac Studio M2 Max (96GB unified memory)
96 GB UNIFIED MEMORY
~$2100 USEDQ3_K_M quant, ~14 tok/s, 16k context with unified memory
- SWEET SPOT
Dual NVIDIA GeForce RTX 3090 (2x24GB) desktop with 64GB RAM
48 GB VRAM PLUS 64 GB RAM
~$1750 USEDQ4_K_M quant with KTransformers offloading, ~22 tok/s, 32k context
- FULL POWER
Workstation with 4x NVIDIA GeForce RTX 4090 24GB
96 GB VRAM
~$7800 NEWFP8 / Q8 quant fully in VRAM, ~45 tok/s, 128k context
LEARN HOW TO SET IT UP
WILL MY BOX RUN QWEN3.5 122B A10B?
The three cards above are researched picks. Search the machine you actually have — we only say yes if it should reply at a usable speed, not just load the weights and crawl.
HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.
WHAT CAN YOU ACTUALLY DO WITH IT?
Everyday jobs on each of those rigs: how long Qwen3.5 122B A10B takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMApple Mac Studio M2 Max (96GB unified memory) | SWEET SPOTDual NVIDIA GeForce RTX 3090 (2x24GB) desktop with 64GB RAM | FULL POWERWorkstation with 4x NVIDIA GeForce RTX 4090 24GB |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~36 S★★★★★ | ~23 S★★★★★ | ~11 S★★★★★ |
| Summarize a documenta long report boiled down to the points that matter | ~1.8 MIN★★★★★ | ~68 S★★★★★ | ~33 S★★★★★ |
| Build a websitea small landing page, markup and styles together | ~9.5 MIN★★★★★ | ~6.1 MIN★★★★★ | ~3 MIN★★★★★ |
| Build a backendan API with routes, storage and tests | ~30 MIN★★★★★ | ~19 MIN★★★★★ | ~9.3 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~18 MIN★★★★★ | ~11 MIN★★★★★ | ~5.6 MIN★★★★★ |
STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.
WHAT THE INTERNET SAYS
People are mostly frustrated with Qwen3.5 122B A10B right now. Biggest praise: speed. Biggest gripe: help with code.
123 POSTS · 34 COMMENTS · REDDIT · HACKER NEWS · LEMMY · GITHUB · DEV FORUMS · DEV.TO · X · BLUESKY · BLOG
PEOPLE LOVE
- Answers come back fast (3 mentions)
PEOPLE COMPLAIN ABOUT
- Coding help disappoints some people (7 mentions)
- Refuses or lectures more than people like (5 mentions)
- Outages, errors, or limits get in the way (3 mentions)
TOPIC BY TOPIC
- Help with codeA SORE SPOT
1 thumbs up · 7 thumbs down
- Saying no too muchA SORE SPOT
0 thumbs up · 5 thumbs down
- SpeedPEOPLE LOVE THIS
3 thumbs up · 0 thumbs down
- Working when you need itA SORE SPOT
0 thumbs up · 3 thumbs down
LOOKING FOR AN ALTERNATIVE?
Other models the crowd has fully reviewed, starting with multimodal models like this one.
Grok 4.6> SOURCES (12)
- HACKER NEWS · 2026-02-28 · 461 UPVOTESQwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers
Qwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers
- REDDIT · 2026-02-24 · 446 UPVOTESQwen3.5 Medium models out now!
Qwen3.5 Medium models out now!
- REDDIT · 2026-08-04 · 267 UPVOTESA llama.cpp PR caches “hot” MoE experts on the GPU — 33 → 56 tok/s reported with 8GB VRAM
A llama.cpp PR caches “hot” MoE experts on the GPU — 33 → 56 tok/s reported with 8GB VRAM
- REDDIT · 2026-02-25 · 174 UPVOTESQwen3.5 Medium GGUFs hit 106K downloads in just 12 hours!
Qwen3.5 Medium GGUFs hit 106K downloads in just 12 hours!
- REDDIT · 2026-02-24 · 80 UPVOTESQwen just released Qwen 3.5 medium model Series: Qwen 3.5 Flash plus 3 more models
Qwen just released Qwen 3.5 medium model Series: Qwen 3.5 Flash plus 3 more models
- REDDIT · 2026-07-09 · 69 UPVOTESIdeogram 4 compared with Krea 2 in natural language prompting
Ideogram 4 compared with Krea 2 in natural language prompting
- HACKER NEWS · 2026-07-11 · 65 UPVOTESFixed three bugs that made Qwen3.5-122B a daily driver on Mac Studio
Fixed three bugs that made Qwen3.5-122B a daily driver on Mac Studio
- REDDIT · 2026-08-04 · 63 UPVOTESDeepseek V4 Flash 2-bit quant is the first model I can run locally that achieves 100% in this SQL benchmark
Deepseek V4 Flash 2-bit quant is the first model I can run locally that achieves 100% in this SQL benchmark
- REDDIT · 2026-07-01 · 51 UPVOTESNVFP4 still isn't faster than FP8 on Blackwell (SM120) - some numbers from Qwen3.6-27B
NVFP4 still isn't faster than FP8 on Blackwell (SM120) - some numbers from Qwen3.6-27B
- REDDIT · 2026-08-02 · 51 UPVOTESAll Qwen model oneshots: 1109 outputs to look at and compare!
All Qwen model oneshots: 1109 outputs to look at and compare!
- REDDIT · 2026-08-02 · 31 UPVOTES[Release] WinterMix — Qwen3.5-122B-A10B in native MLX: an 82 GiB build that beats 94–95 GiB quants, plus a 68 GiB build for agent swarms
[Release] WinterMix — Qwen3.5-122B-A10B in native MLX: an 82 GiB build that beats 94–95 GiB quants, plus a 68 GiB build for agent swarms
- REDDIT · 2026-08-02 · 26 UPVOTESReal-world reality check on Qwen for autonomous coding agents
Real-world reality check on Qwen for autonomous coding agents