Hermes Agent v0.7.0 is out now.
Our headline update:
Memory is now an extensible plugin system. Swap in any backend, or build your own. Built-in memory works out of the box; six third-party providers are ready to go. Pick one with 'hermes memory setup'.
Full changelog below ↓
the people telling you a single 3090 can't ship production quality are not wrong about the ceiling. they're wrong about the conclusion.
most of them prompted a model twice, watched it hallucinate, and made a youtube video titled "local AI is NOT ready." they never iterated. never pushed context. never matched the right model to the right task. never configured a single flag. they gave up where the work starts.
and i never said you'd run frontier agents on a 3090. go back through everything i've posted. what i said is you start there. you learn what your workload actually needs before someone else defines it for you.
because the ones telling you local can't compete have API subscriptions to sell. they have wrappers to maintain. they have inference margins that disappear the moment you run your own hardware. they are not protecting you from frustration. they are protecting their revenue from your independence.
every local AI transition destroys someone's SaaS margin. that's why they fight it.
buy a single GPU. ROME or EPYC board, ECC memory, scalable from day one. cancel every subscription. run local until you hit a wall. and when you hit that wall you will know exactly what cloud compute you actually need instead of what someone told you to buy. use cloud to mine its intelligence then bring it back to your hardware. and scale your GPUs as you go. that's the whole point.
don't build on someone else's thinking. don't store your reasoning on someone else's servers. own the hardware. own the frustration. own what comes after it.
this is what i've been saying since my first post on this account and it's what i will keep saying. there is nothing you can do about it.
@LottoLabs seems kinda clear intelligence has to be local
no limits on intelligence should be the goal (even though they all said "intelligence too cheap to meter")
Qwen 3.5 27b never degrades, never stops running, never has token limits, never refuses, never logs my prompts, never trains on my data, never sells my data, never runs up my credit card
@sudoingX could use their latest infamy to join any of the frontier labs and have a top 0.1% salary
yet they chose, and apparently keep choosing to help opensource. respect
the founder of openclaw joined the company that was founded to make AI open and now charges you per token. and is now telling you open models aren't there yet.
i run qwen 3.5 27b on a single 3090. 50 tok/s. it writes code, handles tool calls, runs agent sessions for hours. the model built a full space shooter, 3,000+ lines, from a single prompt. i published the data.
"open models aren't there yet" is what you say when your harness can't parse tool calls on local models and you blame the model instead of fixing the harness. i have the DMs. people switch from openclaw to hermes agent and their "broken" models suddenly work.
pair a good model with a good harness like hermes agent where parsers are built per model. your data stays on your machine. no API key. 0 subscription. no one training their next model on your thinking.
don't listen to someone with an OpenAI paycheck telling you open source can't do the job. install it. test it yourself. the receipts are on my timeline.
he built a harness that couldn't handle local models and chose the API paycheck over fixing it. that should tell you everything.
the founder of openclaw joined the company that was founded to make AI open and now charges you per token. and is now telling you open models aren't there yet.
i run qwen 3.5 27b on a single 3090. 50 tok/s. it writes code, handles tool calls, runs agent sessions for hours. the model built a full space shooter, 3,000+ lines, from a single prompt. i published the data.
"open models aren't there yet" is what you say when your harness can't parse tool calls on local models and you blame the model instead of fixing the harness. i have the DMs. people switch from openclaw to hermes agent and their "broken" models suddenly work.
pair a good model with a good harness like hermes agent where parsers are built per model. your data stays on your machine. no API key. 0 subscription. no one training their next model on your thinking.
don't listen to someone with an OpenAI paycheck telling you open source can't do the job. install it. test it yourself. the receipts are on my timeline.
he built a harness that couldn't handle local models and chose the API paycheck over fixing it. that should tell you everything.