Joined August 2021
- Tweets 3,093
- Following 280
- Followers 138
- Likes 15,311
257 Photos and videos
Actually, Don't lash out on @FactoryAI. Seems like the BYOK charge is only going to be for Enterprise/Team , Not Individuals. The BYOK usage is pretty high too.
So i think it's fine now, there's been some clear ups.
Factory droid thinks it's okay to charge you money for using your OWN local models if you're on their harness...
They have added a metering system that charges you or stops you from using the models you already pay for if your in factory, this is outrageous, github fix soon
1
2
198
Jun 13
With all of the FactoryAI's `Factory Router` to save money, I wanted to kind of make my own as well, just for fun.. maybe a poor "imitation" of it in a way. And so I did. github.com/definitelynotguru… Made with Cursor & a bunch of GMI (@gmi_cloud) credits.
1
4
79
4/ The actual savings
You don't need 100% accuracy in routing to save money.
Even getting it "mostly right" can cut costs significantly because simple requests dominate usage.
1
22
5/ Building it
Just built this mostly for fun and learning. Cursor CLI. GMI credits (in cursor). A lot of trial and error. Lots of bad ideas deleted. It turns out that “its probs not even difficult to make that on my own” was a bit too hard for me; kudos to @FactoryAI.
29
Jun 13
We need more small models. Especially from @Zai_org and @Kimi_Moonshot. "Flash" like models would be super nice.. alongside something like QAT from Gemma.
18
Jun 13
So far we got:
GLM 5.2
Kimi K2.7 Code
Nice.
Jun 8
it already feels like its been eons since a new model dropped.
@Zai_org when new glm?
@Kimi_Moonshot when new kimi?
@XiaomiMiMo when new mimo?
@Alibaba_Qwen when qwen 3.7 27b, etc. ?
@cursor_ai when /goal?
2
200
Jun 13
GLM 5.2 has a context window of 1M.. interesting, wonder if it hallucinates like deepseek
1
222
What Europe should do right now:
1. Call all the European researchers working on AI and return them back with same salary (or they can stay but switch career).
2. Fill EU places having GPUs with money, and put those people there.
3. AI partnerships with China India.
353
355
3,119
189,193
parth. retweeted
Jun 12
Kimi K2.7 Code is now available on GMI Cloud
10 % off , 72 hours, 30% less token burn than K2.6
Start now
3
6
85
3,121
parth. retweeted
Jun 12
MiniMax M3 is live on GMI
M3 is the first open-weight model combining frontier coding & agent capabilities, 1M-token context, and native multimodal understanding.
Jun 12
MiniMax M3, Open-Weight, Now On Hugging Face , with only ~428B parameters and ~23B activated parameters
Weights:
huggingface.co/MiniMaxAI/Min…
MiniMax Sparse Attention:
huggingface.co/papers/2606.1…
3
3
30
5,545
parth. retweeted
Jun 11
Introducing Automated Security Review in Droid.
5
14
174
29,374
Jun 10
Extremely bullish on Droid & FactoryAI, "Ultracode" is just a worse Droid mission. Everyone should use Mission.
12
374