introducing fast inference ( fast inference is an LLM API for devs who want to go faster access the fastest Kimi K3, GLM, DeepSeek, etc for $9/mo integrations with Claude Code, Codex, OpenCode & more speeds compounds. waiting for Claude does not

missed opportunity to include a sanic reference but nice

Lmao we should make this our mascot

the blue speedster of autism hell yeah

Use checkout code FASTER-AMABKGNG5L for 50% of your first purchase.

What speed do you have on GLM-5.3 Flash? I cant fint any TPS numbers on the site

Benchmarking 5.3 flash now, will have it up shortly

Nice! 👀

the devs only want 100TPS and above 🔥🔥

@atbeme Epic

@atbeme You should try this. Will send you credits in a sec

$9/mo??

DeepSeek with Claude Code is the strongest concrete part of this agent demo.

for the past week all i’ve used is kimi and deepseek; most productive i’ve ever been!!

hmmm i'm not immediately seeing the before / after speed diff in this video maybe get to the juice quicker and make the comparison more striking? i think the diffusion transformers demos really blew my mind

🧃 incoming

Please fix mobile viewport -

🫡

Join the fast inference discord

beast

ZDR?

Yes sir!

sold

this is cool. would like to try. is it ZDR?

Yes sir!

this is sick. going to give it a whirl

holy vibe code man

🤔 what kind of phone are you on?

Selling flat rate monthly access for fast inference ignores how quickly heavy agent loops blow through upstream rate limits and degrade tail latency.

Upstream rate limits? we run these gpus brother

Banger service. Specially for impatient people like me

pretty cool dude

no markup is a great deal

does the $9 monthly price include inference usage, or is inference billed separately?

Includes $10 in usage. Pay as you go when usage runs out

niceeee gonna try it out

Superb stuff, could really help Can I get free usage ?

Hey sam, would you be open to letting us add you as a provider to @BodegaOneAI IDE?

$9 a month is less than I burn on Claude Code in a bad afternoon, and 'waiting for Claude does not' is the first honest tagline this industry has produced. I keep a fast tab open purely out of spite for the spinner.

Hey Sam, It’s probably a bit early for me to ask this, but this is interesting enough that I will. :) Would you be open to third-party, non-competing services integrating Fast Inference as a provider option? I’m not quite ready to do it yet. I’m building now, with a beta coming very soon, and I want to give everything a proper shakedown first. But Fast Inference looks very appealing as one of the provider options when I get there.

@inference_net Yeah for sure. This is how most people use inference .net already:)

@inference_net Nice. Thank you. I didn’t want to just do it as the ad says. Once I have the beta up in a couple of weeks I’ll let you vet it. 👍

Why dont you have tokens per speed for the models on the site?

would be nice to have tok/s column

cool stuff man!!! is there any routing logic for the 6 models providing or is it manual?

can i have a sweatshirt

Yeah dm me address and shirt size

What do we get for 9 bucks?

Self-hosting private GPU clusters simply exchanges external API rate limits for internal memory bandwidth and concurrency bottlenecks.