LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
Qwen3 1.7B
◆ OPEN
by Alibaba
OPEN WEIGHTSLOCALSMALLCHINA-MADEAGENTSUS-MADE
RANKED #231 OF 316 ASSISTANTS · OVERALL #348 OF 493 RANKED · VIBE SCORE 5.6 · 266 VOICES ·
COPY BADGE pastes a README snippet with the rank SVG.
qwen3-asr-1.7b-ja-anime-GGUF on Hugging Face (automatic speech recognition). 3,217 downloads. Open weights for local or hosted use.
HAPPY8.3
16 mentions
SAD6.7
18 mentions
VIBE TRAJECTORY
26 weeks · 320 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 1.7B once and it's yours. No subscription, no rate limits, works offline.
1.7B parameters, a Q4 file is about 1.1 GB, full FP16 is about 3.4 GB
- BARE MINIMUM
Raspberry Pi 5 (8GB)
8 GB RAM
~$80 NEWQ4_K_M, ~12 tok/s, 4k context
- SWEET SPOT
Mac mini M2 (8GB)
8 GB UNIFIED MEMORY
~$450 USEDQ8_0 or FP16, ~65 tok/s, 32k context
- FULL POWER
Nvidia GeForce RTX 3060
12 GB VRAM
~$260 USEDFP16 full precision, ~110 tok/s, 32k full context
LEARN HOW TO SET IT UP
WILL MY BOX RUN QWEN3 1.7B?
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 1.7B takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMRaspberry Pi 5 (8GB) | SWEET SPOTMac mini M2 (8GB) | FULL POWERNvidia GeForce RTX 3060 |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~42 S★★★★★ | ~8 S★★★★★ | ~5 S★★★★★ |
| Summarize a documenta long report boiled down to the points that matter | ~2.1 MIN★★★★★ | ~23 S★★★★★ | ~14 S★★★★★ |
| Build a websitea small landing page, markup and styles together | ~11 MIN★★★★★ | ~2.1 MIN★★★★★ | ~73 S★★★★★ |
| Build a backendan API with routes, storage and tests | ~35 MIN★★★★★ | ~6.4 MIN★★★★★ | ~3.8 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~21 MIN★★★★★ | ~3.8 MIN★★★★★ | ~2.3 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 1.7B right now.
248 POSTS · 18 COMMENTS · GITHUB · DEV FORUMS · REDDIT · LEMMY · DEV.TO
PEOPLE LOVE
- No fan mail yet. The crowd is still warming up.
PEOPLE COMPLAIN ABOUT
- Outages, errors, or limits get in the way (4 mentions)
TOPIC BY TOPIC
- Working when you need itA SORE SPOT
0 thumbs up · 4 thumbs down
LOOKING FOR AN ALTERNATIVE?
Other models the crowd has fully reviewed, starting with text models like this one.
Muse Spark 1.3
DeepSeek V4 Flash> SOURCES (12)
- REDDIT · 2025-06-26 · 197 UPVOTESI tested 10 LLMs locally on my MacBook Air M1 (8GB RAM!) – Here's what actually works-
I tested 10 LLMs locally on my MacBook Air M1 (8GB RAM!) – Here's what actually works-
- REDDIT · 2025-11-25 · 97 UPVOTESFP8 Reinforcement Learning now in Unsloth!
FP8 Reinforcement Learning now in Unsloth!
- REDDIT · 2025-10-14 · 82 UPVOTESI tested local models on 100+ real RAG tasks. Here are the best 1B model picks
I tested local models on 100+ real RAG tasks. Here are the best 1B model picks
- REDDIT · 2026-01-23 · 60 UPVOTESLiquidAI released LFM2.5 Thinking: Runs entirely on-device (phone) with 900MB of memory
LiquidAI released LFM2.5 Thinking: Runs entirely on-device (phone) with 900MB of memory
- REDDIT · 2025-05-23 · 37 UPVOTESTome (open source local LLM + MCP client) now has Windows support!
Tome (open source local LLM + MCP client) now has Windows support!
- REDDIT · 2026-04-23 · 31 UPVOTESReleased a Claude Code skill that fine-tunes a small model from your agent's production traces, end-to-end in one conversation
Released a Claude Code skill that fine-tunes a small model from your agent's production traces, end-to-end in one conversation
- REDDIT · 2025-10-11 · 22 UPVOTES"RLP: Reinforcement as a Pretraining Objective"
"RLP: Reinforcement as a Pretraining Objective"
- REDDIT · 2025-07-23 · 20 UPVOTESMOE Pipeline
MOE Pipeline
- REDDIT · 2025-12-18 · 19 UPVOTESNew functiongemma model: not worth downloading
New functiongemma model: not worth downloading
- REDDIT · 2026-05-19 · 17 UPVOTESI Made LLMs Play Texas Hold’em. The Smallest Model Beat a ~1T Model by Being Too Dumb to Fold
I Made LLMs Play Texas Hold’em. The Smallest Model Beat a ~1T Model by Being Too Dumb to Fold
- REDDIT · 2026-02-10 · 13 UPVOTESNeed help with chunking and embedding strategies for my rag model
Need help with chunking and embedding strategies for my rag model
- GITHUB · 2026-09-28 · 0 UPVOTESThe substrate ladder: three admitted rungs — a small CPU-capable model, Qwen 3.8 27B GSQ-RCO on the 3090 as the always-available floor,…
The substrate ladder: three admitted rungs — a small CPU-capable model, Qwen 3.8 27B GSQ-RCO on the 3090 as the always-available floor, Flash-Next on the Ada instance as capacity — with admission measured by constitutional cells, the depth probe, and a parity…