LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
Qwen3.8 Flash
◆ OPEN
by Alibaba
OPEN WEIGHTSLOCALFAST ANSWERSCODING HELPWRITING HELP
RANKED #3 OF 316 ASSISTANTS · OVERALL #3 OF 493 RANKED · VIBE SCORE 7.4 · 8,180 VOICES ·
COPY BADGE pastes a README snippet with the rank SVG.
A fast, budget-friendly AI for everyday writing and coding help.
HAPPY9.7
508 mentions
SAD5.3
181 mentions
VIBE TRAJECTORY
26 weeks · 8,542 voices in this window
HOLDING UP
Recent vibe is steady. No clear fade in the crowd.
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 Qwen3.8 Flash once and it's yours. No subscription, no rate limits, works offline.
125B MoE with 6B active parameters plus 51B n-gram embeddings, a Q4 quant takes about 94 GB
- BARE MINIMUM
Apple Mac Studio M2 Max (96GB Unified Memory)
96 GB UNIFIED MEMORY
~$2,200 USEDQ2 quant or Q3 with n-gram offloading, ~10 tok/s, 8k context
- SWEET SPOT
Apple Mac Studio M2 Ultra (192GB Unified Memory)
192 GB UNIFIED MEMORY
~$4,500 USEDQ4_K_M quant, ~22 tok/s, 32k context
- FULL POWER
Workstation with 4x NVIDIA GeForce RTX 3090 24GB
96 GB VRAM AND 128 GB SYSTEM RAM
~$3,800 USEDFP8 / NVFP4 quant with host embedding offloading, ~42 tok/s, 64k context
LEARN HOW TO SET IT UP
WILL MY BOX RUN QWEN3.8 FLASH?
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 Qwen3.8 Flash 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 SPOTApple Mac Studio M2 Ultra (192GB Unified Memory) | FULL POWERWorkstation with 4x NVIDIA GeForce RTX 3090 24GB |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~50 S★★★★★ | ~23 S★★★★★ | ~12 S★★★★★ |
| Summarize a documenta long report boiled down to the points that matter | ~2.5 MIN★★★★★ | ~68 S★★★★★ | ~36 S★★★★★ |
| Build a websitea small landing page, markup and styles together | ~13 MIN★★★★★ | ~6.1 MIN★★★★★ | ~3.2 MIN★★★★★ |
| Build a backendan API with routes, storage and tests | ~42 MIN★★★★★ | ~19 MIN★★★★★ | ~9.9 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~25 MIN★★★★★ | ~11 MIN★★★★★ | ~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
Users love Qwen3.8 Flash for its quick responses, solid writing, and strong coding help at a great value. However, people are frustrated by frequent crashes, strict refusals, and occasional made-up information.
2628 POSTS · 5552 COMMENTS · GITHUB · REDDIT · HACKER NEWS · BLUESKY · LEMMY · BLOG · DEV.TO · DEV FORUMS
PEOPLE LOVE
- Popular for coding help (11 mentions)
- Answers come back fast (9 mentions)
- People feel it is worth the money (7 mentions)
- Good at writing and wording things (5 mentions)
PEOPLE COMPLAIN ABOUT
- Refuses or lectures more than people like (22 mentions)
- Outages, errors, or limits get in the way (16 mentions)
- Sometimes makes things up (6 mentions)
TOPIC BY TOPIC
- Saying no too muchA SORE SPOT
0 thumbs up · 22 thumbs down
- Working when you need itA SORE SPOT
2 thumbs up · 16 thumbs down
- Help with codePEOPLE LOVE THIS
11 thumbs up · 3 thumbs down
- SpeedPEOPLE LOVE THIS
9 thumbs up · 3 thumbs down
- PriceMOSTLY POSITIVE
7 thumbs up · 3 thumbs down
- Getting facts rightA SORE SPOT
2 thumbs up · 6 thumbs down
- WritingPEOPLE LOVE THIS
5 thumbs up · 1 thumbs down
- Long conversationsPEOPLE LOVE THIS
3 thumbs up · 0 thumbs down
IN THEIR OWN WORDS
- FROM GITHUB >
“Add the qwen3.8-flash ceiling arm the release gate refused v0.13.9 over.”
- FROM GITHUB >
“Qwen 3.8 Flash Next (qwen4exp): fatal broadcastshapes crash at step 2 of any tool run — QSA indexer budget vs KV length.”
- FROM GITHUB >
“llama-cpp: qwen3.8-flash crash-loops on load (qwen4exp warmup decode abort).”
- FROM REDDIT >
“Prefill is ~150 t/s. The model itself is crazy smart - better than the iq4-xs quant of Qwen3.8-Flash-Next I've been running for certain - but it likes to think A LOT and with the already slow average of 9t/s decode I won't be using it for much outside of chatting even if it is the new best model on my machine.”
LOOKING FOR AN ALTERNATIVE?
Checked picks first: tags say why you would switch, stars say how fully each one stands in for Qwen3.8 Flash.
Gemini 3.7 Flash
Grok 4.6> SOURCES (12)
- REDDIT · 2026-08-25 · 1029 UPVOTESQwen3.8-Flash-Next announced
Qwen3.8-Flash-Next announced
- REDDIT · 2026-08-26 · 583 UPVOTESRun Qwen3.8-Flash-Next locally! 🔥
Run Qwen3.8-Flash-Next locally! 🔥
- REDDIT · 2026-08-26 · 505 UPVOTESQwen releases Qwen3.8-Flash-Next!
Qwen releases Qwen3.8-Flash-Next!
- REDDIT · 2026-08-27 · 365 UPVOTESAll Qwen3.8-Flash-Next quants are up!
All Qwen3.8-Flash-Next quants are up!
- REDDIT · 2026-09-02 · 281 UPVOTESRun Qwen3.8-Flash-Next GGUF 1.7x Faster!
Run Qwen3.8-Flash-Next GGUF 1.7x Faster!
- REDDIT · 2026-08-27 · 246 UPVOTESQwen3.8-Flash-Next now available in Unsloth!
Qwen3.8-Flash-Next now available in Unsloth!
- REDDIT · 2026-10-17 · 220 UPVOTESWhy not run Qwen 3.8 Flash Next with the full n-gram-table?
Why not run Qwen 3.8 Flash Next with the full n-gram-table?
- REDDIT · 2026-08-26 · 185 UPVOTESQwen 3.8 Flash Next: Beating DS V4 Flash at half the parameters, stronger than Opus 4.6
Qwen 3.8 Flash Next: Beating DS V4 Flash at half the parameters, stronger than Opus 4.6
- REDDIT · 2026-08-27 · 178 UPVOTESQwen3.8-Flash-Next 176B on a 16GB card: yes it works, here's how
Qwen3.8-Flash-Next 176B on a 16GB card: yes it works, here's how
- REDDIT · 2026-10-17 · 178 UPVOTESXolotl (Strata based) - DeepSeek-V4.1-Flash (552B) on a single 4090 at ~27 tok/s, Uncensored 1017 quant coming in a couple of days
Xolotl (Strata based) - DeepSeek-V4.1-Flash (552B) on a single 4090 at ~27 tok/s, Uncensored 1017 quant coming in a couple of days
- REDDIT · 2026-09-02 · 173 UPVOTESNew Unsloth Release - much faster performance + new model support
New Unsloth Release - much faster performance + new model support
- REDDIT · 2026-10-19 · 153 UPVOTESISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-Coder-GGUF updated
ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-Coder-GGUF updated