Earlier quoted context omitted.
Isn't Kimi K2.6 natively INT4?
I don't think any models are natively INT4? I wouldn't see the point to nerf the model out-of-the-box.
The only model I have seen like that is GPT OSS, natively quantized to MXFP4.
Mistral Medium 3.5
241–248 of 248 posts
Re: Mistral Medium 3.5
#242Earlier quoted context omitted.
> I've seen the idea that GPT-2 not being released was marketing hype at least 6 times since Mythos was shared. That's not what I am saying. It's not that GPT-2 not being released was marketing hype, it's that OpenAI themselves claiming it's too dangerous to release specifically, implying it's close to AGI, (or something like that), was marketing hype.
That may sound more defensible to you, but its even more detached from reality. I feel very old right now because I actually read the thing at the time, but setting that aside, do you really think anyone thought or said GPT-2 was AGI ? I don't think you do. I only mention reading it because that would clear it up, and you seem interested, and your parenthetical indicates A) you're aware you're claiming something a bi…
> do you really think anyone thought or said GPT-2 was AGI?
> I don't think you do.
I don't think they did. I think the marketing around it was such as to imply to the general public that it's not that far from AGI/let their imagination run wild, because it's useful marketing.
Same way as Apple's marketing around the iPad was that it's the 'Super. (full stop) (space) Computer. (full stop)' They never say it's any kind of a 'supercomputer'. But they know how that is going to be interpreted by many. It's intentional marketing.
Same with OpenAI.
I do think you're playing obtuse at this point, if you don't get that.
Re: Mistral Medium 3.5
#243Earlier quoted context omitted.
That may sound more defensible to you, but its even more detached from reality. I feel very old right now because I actually read the thing at the time, but setting that aside, do you really think anyone thought or said GPT-2 was AGI ? I don't think you do. I only mention reading it because that would clear it up, and you seem interested, and your parenthetical indicates A) you're aware you're claiming something a bi…
> do you really think anyone thought or said GPT-2 was AGI? > I don't think you do. I don't think they did. I think the marketing around it was such as to imply to the general public that it's not that far from AGI/let their imagination run wild, because it's useful marketing. Same way as Apple's marketing around the iPad was that it's the 'Super. (full stop) (space) Computer. (full stop)' They never say it's any kin…
I am not playing obtuse, and note “do you really think” is different from accusing someone of playing a character on purpose to gaslight you, especially when the “do you really think they said” is stapled to a long, kind, explication that your parenthetical says straight out you don’t know what was said, you’re guessing.
Your attempt to recount GPT-2 and what was said about it won’t make any sense to anyone who was present at the time, and you’ve circled back to root, “marketing”, again, for a company that was selling nothing and wouldn’t for years and wasn’t raising and wasn’t even mentioning AGI in connection with GPT2. It was just a text generator back then, though a curious one worth noting. It’s genuinely awe inspiring to see someone arguing gpt-2 was near AGI and that anyone at the time thought so or said so.
It’s not worth engaging further, especially with you starting to get nasty and project what’s in my head. Have a good day.
Re: Mistral Medium 3.5
#244I like the idea of Mistral, but the last time I evaluated Mistral Vibe it was really nice for $15/month but not as effective as Gemini Plus with AntiGravity and gemini-cli. I am currently running Gemini Ultra on a 3 month 'special deal' and AntiGravity with Opus 4.7 tokens is pretty much fantastic. That said, when I stop spending money on Gemini Ultra, I will give Mistral Vibe another 1-month test. I like the entire…
How do you feel about the responsiveness of gemini-cli? I tried it on a paid plan and the 10-minute hang-ups (per step, not the whole plan execution) really break the illusion of performance gains, unless you run it in the background and do something else in the meantime. It's more noticeable when Americans are awake.
it is usually fast, but if gemini-cli or any other coding agent is sluggish I quit using it for a while.
Re: Mistral Medium 3.5
#245Earlier quoted context omitted.
Recent models support multi-token prediction, which can guess multiple future tokens in a single decode step (using some subset of the model itself, not a separate drafting model) and then verify them all at once. It's an emerging feature still (not widely supported) and it's only useful for speeding up highly predictable token runs, but it's one way to do better in practice than the common-sense theoretical limit mi…
If Mistral Medium 3.5 supports it, that might get it to 10 t/s. It will still be fairly slow.
[dead]
Re: Mistral Medium 3.5
#246Earlier quoted context omitted.
They did credit it back to him. There's a comment in the linked issue.
Where? Just searched the entire thread for both the word "refund" and the word "credit" and I'm seeing nothing about credit being issued. Also what's with @sasha-id talking to himself? Looks weird as all get out.
Re: Mistral Medium 3.5
#247Earlier quoted context omitted.
> Pre-agent, there wasn't always an obvious difference between models. Various models had their charms. Nowadays, I don't want to entertain anything less than the frontier models. This is a very naive and misguided opinion. In most tasks, including complex coding tasks, you can hardly tell the difference between a frontier model and something like GPT4.1. You need to really focus on areas such as context window, tool…
> You need to really focus on areas such as context window, tool calling and specific aspects of reasoning steps to start noticing differences. This is like saying "the current models and the old models are the same if you ignore every important advance they've made"
> This is like saying "the current models and the old models are the same if you ignore every important advance they've made"
Please go ahead and list the single most important advance a frontier model has over, say, gpt4.1.
Reasoning is one of the main features, and in practice all it does is waste compute to rewrite your original prompt. See how GPT 5.4 burns through compute by running additional prompts where it acts like your own interpreter, with lengthy reasoning prompts on "the user is asking for (...)" as if you are completely unable to stitch together a usable prompt. That's your frontier model.
Re: Mistral Medium 3.5
#248Earlier quoted context omitted.
Keep this in mind next time you hear someone talking about "removing the human in the loop". Anthropic apparently won't take responsibility for issues their own systems handling billing cause. You think they'll take responsibility in your system when a bug in their models can be demonstrated as the cause?
> You think they'll take responsibility in your system when a bug in their models can be demonstrated as the cause? Flag on the play: AI doesn’t replace responsibility for your commits. It doesn’t matter what promises a service makes, what you say is valid code is still on you. Act accordingly.
The issue is less what's in your commit and more if you're using these models as a foundation for some other service.
I know this is a rather hackneyed example, but if a customer service agent model were to call a customer a racial slur, that's not the software surrounding the agent, it's the agent's model.