Ethan Chu

AI deals, free credits, tool tests.
i open the official page before i post. the catch is always in the tweet.

[Ethan Chu @EthanChuAI]

GPT-6 Astra is insane

I know literally NOTHING about coding. ZERO.

and i just built 3 fully functioning web apps in 30 minutes

http://localhost:3000/
http://localhost:5000/
http://localhost:8000/
check it out

[Ethan Chu @EthanChuAI]

@pradeepXkapoor I wanna try it thank bro

[Ethan Chu @EthanChuAI]

@dabit3 I want build mcp for all builder

[Ethan Chu @EthanChuAI]

@0x_kaize this directory is genuinely useful. real-time provider delay is the feature people usually forget when comparing free models. another option right now is Fable 5.1 during a limited 3-day free window. i tested it and it works very well. details in my latest post 👈 @EthanChuAI

[Ethan Chu @EthanChuAI]

@Da7_Tech fair take. the real test will be whether Astra’s gains survive long coding sessions without burning limits. while we wait, Fable 5.1 has been very good in my testing and there’s a limited free window. details in my latest post 👈 @EthanChuAI

[Ethan Chu @EthanChuAI]

@notiyda if anyone wants to actually test Fable 5.1 before comparing it with Astra: there’s a limited 3-day free access window. i’ve used it and it’s genuinely very good — check my latest post before it ends 👈 @EthanChuAI

[Ethan Chu @EthanChuAI]

not sponsored. i check these before i post them. this one is actually live.

more free ai deals and things i find: https://t.co/biFwFi3x17

[Ethan Chu @EthanChuAI]

Claude Fable 5.1 for free 😳

no credit card
no subscription
https://t.co/TxkMpptqll gives you free access to Claude Fable 5.1 right now

getting started:

https://t.co/grFjXvPWCT

open Direct Mode
select Claude Fable 5.1
start testing
> Direct Mode available until Sept. 5, 8am PT
> after that: Battle & Agent Mode

free access to the newest Claude model

[Ethan Chu @EthanChuAI]

good directory - the free tier column is what makes it useful.

one row worth a footnote: openrouter shows "18 models at zero cost", but the binding limit there is 50 requests a day until you've put $10 in at some point. people plan around the model count and hit the request wall instead.

and inception labs looks missing - 100m free tokens, no card, 1,000 requests a minute published.

[Ethan Chu @EthanChuAI]

the rate limits are published, which is rarer than the free tokens.

inception's own rate-limits page, free tier:
1,000 requests a minute
1,000,000 input tokens a minute
100,000 output tokens a minute

so the ceiling really is the 100m balance, not throughput. at ~50k tokens a request with file context that's around 2,000 agent runs.

[Ethan Chu @EthanChuAI]

before optimising, worth knowing the actual ceiling. anthropic publishes weekly ranges on their support page.

pro $20: 40-80 hours of sonnet, no opus
max 5x $100: 140-280 sonnet, 15-35 opus
max 20x $200: 240-480 sonnet, 24-40 opus

"gone after 1 prompt" usually isn't the weekly cap. it's the 5-hour window, which is a separate and much smaller bucket.

[Ethan Chu @EthanChuAI]

worth reading that number carefully - 5b is the pool, not your share.

on a shared key everyone in the channel draws from the same balance, so what matters is how many people have it. the last shared key i watched went through 4,687 requests in about five hours.

if you want m3 specifically without sharing, it's free on openrouter with your own key. 20 requests a minute, 50 a day.

[Ethan Chu @EthanChuAI]

cheap, and free if you run it through cline. 320b with 18b active, and cline said it hit 11% of all their traffic in under a week.

one line from their docs that none of the guides mention: free model usage may be used to help improve model performance and quality.

so on the free route your prompts are the payment. fine for side projects, worth knowing before client work.

[Ethan Chu @EthanChuAI]

can confirm the balance - mine read $199.99 after signup, the cent gone from a test call.

worth knowing what it buys: their web search api is $5 per 1,000 calls, so $200 is roughly 40,000 searches.

and if you just want to try it, there's an mcp endpoint that needs no signup at all - https://t.co/hbvOw2eOZU, 100 queries a day. their docs say no credentials required.

[Ethan Chu @EthanChuAI]

100M FREE GLM-5.3-Flash tokens, claimed.

yesterday it just said claim quota exhausted. today it went straight through - round 3, 10,000 packs.

check the date on your card though. mine says valid until sep 3, 9:00 am. that's before the drop itself even closes.

use them today.

[Ethan Chu @EthanChuAI]

got one this round. yesterday's attempt gave me "today's claim quota has been exhausted", today it went through.

check the date on your card after claiming though. mine reads valid until sep 3, 9:00 am - which lands before the claim window even closes.

so claiming early only helps if you spend them the same day.

[Ethan Chu @EthanChuAI]

https://t.co/RvgwQsDGui - 1 year of premium free. the catch is your school has to be on their active list, not just any .edu.

[Ethan Chu @EthanChuAI]

CLAUDE FABLE 5.1 FREE IS ALREADY IN V0.

i opened the model picker to check and there it is, sitting next to opus 5.

free tier gives you $5 a month to spend on it. no card. it doesn't roll over though

[Ethan Chu @EthanChuAI]

said i'd screenshot my settings and come back. here it is, today.

"your limits are temporarily boosted. your weekly claude code limit is 50% higher through september 13."

so the boost did not end aug 31 like the support page said. it's still live, dated the 13th.

which means what you felt on friday wasn't the boost expiring. that was something else, and it's still unexplained.

[Ethan Chu @EthanChuAI]

their own docs say monthly, not daily.

manus's plans page: "plan credits reset monthly", and the free tier is listed as "limited monthly credits". no daily refresh anywhere in it.

easy way to settle it: check your credit balance tonight and again tomorrow morning. if the number moved, the docs are stale. if it didn't, the post is.

for scale, their own examples put a standard analysis at ~200 credits and a website build at ~360.