Claude Opus 4.8

RIP Fable 5 - You will be missed June 9, 2026 – June 12, 2026.

Wait, what? Claude can't generate images, but can generate complete game assets?

Sir I respectfully disagree. That museum build is definitely worth 64 dollars! that was amazing!

I hope you posted that fable 5 museum up, cause that is some great stuff.

One area where I noticed a remarkable difference is that it has gotten a lot less willing to blindly "bend" towards your PoV. If I state a provably false assumption as truth, 4.7 would often resort to the famous "You are absolutely brilliant for having these thoughts (sycophantic response) because of A and B -- but I want to push back (mitigating response) on C". Then, if you kept pressing, it would just take over the whole false assumption as truth for the rest of the conversation. Opus 4.8 will more often solidly dig its heels into the ground, especially when it's CoT provides evidence that the initial prompt has serious logical flaws in it. It will really call you out, debate and attempt to change your mind. It will still comply if you explicitly demand it, but it won't just let YOUR lies/errors slip past with a small side note anymore either. I can understand not everyone would call this an improvement, but so far, for me, it has been a positive.

Absolutely mental! Fable is really just a whole new level. They weren't kidding.

Acho que o claude faltou colocar no max, e no modo thinking também para melhores resultados.

Times like this remind me that despite GLM and Codex and other models being hyped up as Claude Opus 4.8 replacements, I still would not trust them with my most important work. For example, right now I'm working on a huge refactoring project, and even Opus has struggled with it after several days. I cannot even imagine how GLM, Codex, or other models would handle this. So the only option for me is to wait until this outage is over. And it's not like open models are cheap to run even as alternatives. For example, with my $100/mo subscription for Claude Code, I often burn more than $100 a day several times a week. But if I were to use the API of GLM, it would be about $300.

17:15 not worth it? are you kidding me? maybe the specific build is not really worth it, but the quality and amount of the code for that price is absolutely amazing..what would it have cost done by a human? how much would it cost if you gave opus 4.8 or the older models a few more prompts to fix it and get it to the same level? -a codex user

Just curious how the router decides on which model to use. When I use Claude Code, I often ask Claude Code to decide itself if it should spawn a sub-agent to downgrade or upgrade the model. Claude Code is smart to know how much context and cache it has and will decide if it should use sub-agent with a lesser model (sometimes it costs more to re-fetch tokens with a Sonnet sub-agent if the parent agent already has the context).

Claude Fable 5 has an output price of $50 per 1m tokens, and GLM-5 has an output price of $4.4, making: GLM-5 11.36x cheaper than Fable 5

propuesta de titulo para el video: Probé Claude Opus 4.8 y tengo DEUDAS

3-way artificial unintelligence clash

I think there is a serious question where the graphics assets are from since claude has no image generation. Did it buy textures? I am using Fable and it is very clear it cannot generate textures you have to supply the textures. What is the language, platform or engine used for the game? Lots of questions and it got the least time. The video spent more time on images of candles. Something is odd about that.

I'm glad there are more attempts at solving model routing, as costs (at API rates) has really become an issue. Some feedback: 1. Reiterate the cache issue from other comments already here. there is a lot of optimisation in harnesses around caching and a proxy model blows that up 2. Coding agents are model aware - they already route code discovery to mini / flash models, planning to heavy models, workflow design to ultra, implementation to mid / high etc. They know when they're exploring, planning, implementing, reviewing etc. and which model class to select and when it fails. With a proxy you're breaking this control loop and feedback. It doesn't know, for ex. that it just attempted with deepseek v4 and it failed, lets try Opus? 3. How are you going to RL improvements and prevent the router becoming stale? You only have access to your own internal prompts and ~thousands of samples. This is RL'd on one orgs codebase. There are going to be a lot of prompts you haven't seen before and have no insight to on how to route correctly, and you have no insight into users HF to improve your own model. Orgs aren't going to share their traces with you, so you need other sources to train on and improve There are also new model releases every week that you need to keep up with - whats the story going to be here 4. Publish evals by running terminalbench / deepswe bench. Show us the performance / cost / time chart vs the other agent and model sets. If you can show gains there, you have a very simple value prop to sell where you can charge for a % of the saved costs

It depends what you’re doing. I have no problem using an open source model to build a static website or other simple tasks. And of course, open source is catching up and the things you can trust it with grows every month. Not to mention the way the proprietary models patronize you if they think you’re up to no good. The other day I was trying to use Claude to transcribe a song on YouTube to sheet music; the link was broken and it had not been available for purchase for years. Of course I first had to prove to Claude that I was not just being cheap and trying to get around paying the $5 for the download, which I would have had no problem with. “I am going to be up front with you —- my research raises some red flags with your initial premise that the sheet music is no longer available…” I think in the next 5-10 years inference costs will come down substantially and the open source models will get so advanced that only someone in an extremely niche, cutting edge field will need a frontier model. Everyone else will have a dedicated box somewhere on their network that runs their LLM of choice. No tokens, none of your data getting sent to a third party, no arguing with it over whether or not a link is actually broken or if you’re just trying to be cheap. I think OpenAI and Anthropic realize this which is the reason they’re in a rush to go public.

A lot of the perception of open source models being garbage is that they're still using the same piss-poor sampling algorithms that OpenAI/Anthropic force on their users, i.e. Top-p, top-k. These lead to small accumulation of sampling errors which makes it all but inevitable that open source models will shit the bed by the 200K token mark or even sooner. If you set your opencode to use a good sampling algorithm, such as min_p or top-n sigma (llamacpp supports both), you'll find that at least for long running tasks, your model gets a lot better. It won't make GLM as good as Opus 4.8, but it will stop the feeling of "brain damage" from running open source models at the edge of their context windows. And yes, there is an upcoming (hopefully NeurIPS) paper titled "Long Context Generation is a Sampling Problem" for more details about this. Give it two months and it'll be on Arxiv one way or another.

Opus 4.8 is nearly unusable at the moment. Making extremely obvious errors, failing to debug issues even after being prompted with the exact line of code where the problem is. It should be illegal not to disclose serious degradations of performance while still charging full price. It may just be my region's datacentres degraded or something. The company is so opaque about what's happening, it's impossible to know what is actually going on. The optimistic take is they're preparing to relaunch Fable 5 in the next day or so and siphoning off power for that. Perhaps they're doing a snap retraining to strip out any cybersecurity capability.

This would have been neat back when I could afford enough tokens to even set it up properly. Now I’ve had to increase my GH Copilot subscription just to cover the bare minimum updates to a few websites every month, and I no longer do any test driving, or even recreational coding projects. I don’t have hundreds of dollars a month to plow into these products, so I’m rationing use, looking for better local options, and being much more discerning about where these tools actually save time. Precarious time to be alive…

It's rather hard to do at the proxy level with agentic coding, such as Claude Code or similar. These are long-chained sessions of tool use that heavily rely on prompt caching. Changing mid-flight is costly. It looks like much more context is required to decide on the best model (e.g., summarizing logs might use a cheap model, whereas you likely want Opus/Mythos/GPT 5.6 to debug multithreading logic). In an agentic system, a decision about the model may be embedded in the decision to orchestrate the model.