LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
gpt oss 20B
◆ OPEN
by OpenAI
OPEN WEIGHTSLOCALAPIPAIDFREE TO TRYCLOSED
RANKED #118 OF 228 ASSISTANTS · OVERALL #194 OF 6,618 · VIBE SCORE 5.8 · 400 VOICES
gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for...
HAPPY8.1
53 mentions
SAD6.9
36 mentions
VIBE TRAJECTORY
26 weeks · 371 voices
QUIETLY DEGRADING
Recent vibe and chatter are both sliding.
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 gpt oss 20B once and it's yours. No subscription, no rate limits, works offline.
21B parameters (MoE, 3.6B active), native MXFP4/Q4 file is about 14 GB
- BARE MINIMUM
NVIDIA RTX 4060 Ti 16GB
16 GB VRAM
~$370 USEDMXFP4/Q4, ~15-20 tok/s, 8k context
- SWEET SPOT
NVIDIA RTX 3090 24GB
24 GB VRAM
~$850 USEDMXFP4/Q4 or Q8, ~30-40 tok/s, 32k context
- FULL POWER
Mac Studio M2 Max (64GB RAM)
64 GB UNIFIED MEMORY
~$1,700 USEDUnquantized FP16 or Q8, ~25 tok/s, full 128k context
LEARN HOW TO SET IT UP
WILL MY BOX RUN GPT OSS 20B?
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 gpt oss 20B takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMNVIDIA RTX 4060 Ti 16GB | SWEET SPOTNVIDIA RTX 3090 24GB | FULL POWERMac Studio M2 Max (64GB RAM) |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~29 S★★★★★ | ~14 S★★★★★ | ~20 S★★★★★ |
| Summarize a documenta long report boiled down to the points that matter | ~86 S★★★★★ | ~43 S★★★★★ | ~60 S★★★★★ |
| Build a websitea small landing page, markup and styles together | ~7.6 MIN★★★★★ | ~3.8 MIN★★★★★ | ~5.3 MIN★★★★★ |
| Build a backendan API with routes, storage and tests | ~24 MIN★★★★★ | ~12 MIN★★★★★ | ~17 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~14 MIN★★★★★ | ~7.1 MIN★★★★★ | ~10 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
The internet is split on gpt oss 20B. Biggest praise: help with code. Biggest gripe: working when you need it.
323 POSTS · 77 COMMENTS · REDDIT · STACK OVERFLOW · DEV FORUMS · GITHUB · BLUESKY · DEV.TO · LEMMY · X · BLOG · HACKER NEWS
PEOPLE LOVE
- People feel it is worth the money (3 mentions)
- Popular for coding help (3 mentions)
PEOPLE COMPLAIN ABOUT
- Outages, errors, or limits get in the way (4 mentions)
TOPIC BY TOPIC
- PriceMOSTLY POSITIVE
3 thumbs up · 2 thumbs down
- Help with codePEOPLE LOVE THIS
3 thumbs up · 1 thumbs down
- Working when you need itA SORE SPOT
0 thumbs up · 4 thumbs down
LOOKING FOR AN ALTERNATIVE?
Checked picks first: tags say why you would switch, stars say how fully each one stands in for gpt oss 20B.
Muse Glimmer 30B
Mistral Small
Grok 4.6
Qwen3.8 27B> SOURCES (12)
- REDDIT · 2026-02-17 · 587 UPVOTESCar Wash Test on 53 leading models: “I want to wash my car. The car wash is 50 meters away. Should I walk or drive?”
Car Wash Test on 53 leading models: “I want to wash my car. The car wash is 50 meters away. Should I walk or drive?”
- REDDIT · 2025-08-05 · 485 UPVOTESOpen AI GPT-OSS:20b is bullshit
Open AI GPT-OSS:20b is bullshit
- REDDIT · 2025-08-10 · 250 UPVOTESI ran OpenAI’s GPT-OSS 20B locally on a 16GB Mac with Ollama — setup, gotchas, and mini demo
I ran OpenAI’s GPT-OSS 20B locally on a 16GB Mac with Ollama — setup, gotchas, and mini demo
- REDDIT · 2025-08-24 · 206 UPVOTESWhy doesn't he just creat an ai powered suggestion box that scrubs and categorizes suggestions?
Why doesn't he just creat an ai powered suggestion box that scrubs and categorizes suggestions?
- REDDIT · 2026-02-01 · 150 UPVOTESsome uncensored models
some uncensored models
- REDDIT · 2026-08-17 · 139 UPVOTESQwen 3.8 27B Overthinking, It has to be done, it has to be overthinking to punch Opus 4.6
Qwen 3.8 27B Overthinking, It has to be done, it has to be overthinking to punch Opus 4.6
- REDDIT · 2026-02-12 · 136 UPVOTESJust try gpt-oss:20b
Just try gpt-oss:20b
- REDDIT · 2026-04-07 · 135 UPVOTESI ran Gemma 4 26B vs Qwen 3.5 27B across 18 real local business tests on my RTX 4090. Gemma won 13 to 5.
I ran Gemma 4 26B vs Qwen 3.5 27B across 18 real local business tests on my RTX 4090. Gemma won 13 to 5.
- REDDIT · 2025-08-06 · 126 UPVOTESYou can now run OpenAI's gpt-oss model at home!
You can now run OpenAI's gpt-oss model at home!
- REDDIT · 2025-08-07 · 124 UPVOTESBest models under 16GB
Best models under 16GB
- REDDIT · 2026-04-08 · 112 UPVOTESM5 Max 128GB, 17 models, 23 prompts: Qwen 3.5 122B is still a local king
M5 Max 128GB, 17 models, 23 prompts: Qwen 3.5 122B is still a local king
- REDDIT · 2025-09-29 · 105 UPVOTESUncensored GPT-OSS-20B
Uncensored GPT-OSS-20B