@cheeez42

Local LLM benchmarks on consumer GPUs | tests, reproducible configs, practical verdicts

wumbolabs.dev
Joined December 2018
  • Tweets8.4K
  • Following650
  • Followers950
996Media
r9700s are now 2400+…..FUCK
1
5
153
yea, like justin said adjust your strategy…use applystead.com

Free AI resume tailoring, and help applying to jobs

Tailor your resume to any job with free AI, no account needed. Export clean PDFs and check keyword coverage. Premium has a 7-day free trial.

applystead.com
Justin Hammon
@justinhammon_23h
You've sent 80 applications and heard back from 4. The reason: Not on you. The average job search duration just hit a 4-year high, per BLS data. The market is frozen. Adjust your strategy. Don't blame your resume.
1
143
yss, please push 800+w on your 5090 with a 600w cable that already melts at stock power settings.
SuSu_酥酥👅
@NFT_ChenOct 25
🚨预警！英伟达RTX 5090破解方案应该很快就出来了！中国工厂已经在拆5090做AI板，破解一出需求会更炸！ RTX系列AI算力大解锁，老黄刀法持续被破解，已知软件解锁，驱动前覆盖寄存器，无硬件改动： 
• 2080 Ti：59.9→111 TFLOPS（~1.9×）
• 3090：73→122.8 TFLOPS（~1.7×）
• 3090 Ti：87.5→157.4 TFLOPS（~1.8×）
• 4090：165.6→331.6 TFLOPS（~2×） 预测5090：目前FP32-acc路径锁在~210 TFLOPS，一旦破解有望冲到~420+ TFLOPS。 再叠加现有800W+ BIOS/功耗解锁，本地AI直接起飞。 #RTX5090 #AI解锁 #GPU破解 #NVIDIA #Blackwell #本地AI #TensorCore #黄仁勋
2
7
548
terra, my beloved.
OpenAI
@OpenAI23h
Replying to @OpenAI
GPT-6.1 Terra is rolling out now and is available to all Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex. Developers can also access it through the OpenAI API as gpt-6.1-terra. openai.com/index/introducing…

Introducing GPT-6.1 Terra

Delivering fast and capable computer use at scale

openai.com
1
4
306
oof
下田悠太 | Yuta Shimoda🐡@simo4mo314Oct 26
Local LLM構築のため、はじめての自作PC組み立てに挑戦🔥
4
1
21
641
tern for ios is awesome. now i can keep the clankers working while im at work.
1
2
127
full qwen3.8-flash-nexy iq3_s getting around 35tok/s locally.....fuck ill take it.
5
163
*we have deepseek at home*
cheez | Local LLM Benchmarks
@cheeez42Oct 26
going to give deepseek v4 flash 1017 with strata and then glm5.3-flash with project maya a try on wumbojets and see how it flys...this should at least tell me how much debt i am going to amass in the next week or two.
2
9
224
going to give deepseek v4 flash 1017 with strata and then glm5.3-flash with project maya a try on wumbojets and see how it flys...this should at least tell me how much debt i am going to amass in the next week or two.
1
2
356
anyone ever run imto a NV_ERR_NO_MEMORY (0x51) with any nvidia blackwell cards. seen it pop up a couple times since reinstalling my 5070 and cant narrow down the cause.
45
the only bench that matters is runebench.
27
hell yea. Niko continues to crush.
Niko Veit
@coldnikoOct 25
We are getting closer to our goal! More to follow soon 🫡
2
138
just do both
wolfie
@wolfie_Oct 25
fuck fake it til you make it. larp it til you are it
5
284
so the ideas i have come up with so far: a. rob bank b. drive to the city and stand on the corner c. spend my own money d. find a wealthy sugar something.... 🤔
cheez | Local LLM Benchmarks
@cheeez42Oct 24
gonna take a nap and dream about ways to get more ram…
1
5
151
wolfie approved
wolfie
@wolfie_Oct 23
Replying to @cheeez42
hell yeah king
1
4
225
gonna take a nap and dream about ways to get more ram…
1
1
8
436
dear santa, please put at least 96gb of ddr5 under the tree this year so i can deepseek at home.
1
3
92
lets goooo. we might actually hit the 1k by the end of the year. ~55 more followers with ~70 days until the year ends. that seems pretty reasonable. if you are new here. welcome and thank you. im cheez. restaurant work by day, larping with AI in my free time.
1
12
278
let me find these guys. idc if they are stolen. lemme get a 5090 or 10.
VideoCardz.com
@VideoCardzOct 22
Someone tried to sell an RTX 5090 from a stolen $470K shipment online, with its serial number visible 🔗videocardz.com/newz/someone-…
3
185
Load more