Anthropic's New Opus 5 Model Rivals Fable 5 For Half the Price
- Reference: 0184656206
- News link: https://slashdot.org/story/26/07/24/1853236/anthropics-new-opus-5-model-rivals-fable-5-for-half-the-price
- Source link:
> One important distinction between Opus 5 and Mythos, which is currently only available to a limited number of vetted organizations through [3]Anthropic's Project Glasswing initiative , is that the company has specifically avoided training the new model on cyber-related tasks. Due to more its powerful capabilities, Opus 5 is broadly better at those tasks than its predecessor, making it more useful for finding cybersecurity vulnerabilities, but the company says Opus 5 is "substantially behind" its flagship model at exploiting those vulnerabilities.
>
> [...] Anthropic says "Claude Opus 5's safeguards are designed to allow beneficial uses of the model in both cybersecurity and biology." The company has strengthened some of the model's cyber-related guardrails, but notes it did so along a "narrow range" of specific tasks. "Based on our testing, we expect the classifiers to intervene around 85 percent less often than they do for Fable 5," Anthropic said. With Opus 5, Anthropic also isn't including it in its recently announced 30-day data retention policy, which the company introduced alongside Fable and Mythos 5.
>
> As for pricing. Anthropic says API costs for Opus 5 will remain at $5 per one million input tokens and $25 per one million output tokens. Anthropic has also added an "effort" menu for Opus that users can tweak to tell the model whether they want it to be more thorough or fast and efficient to conserve tokens.
[1] https://www.anthropic.com/news/claude-opus-5
[2] https://www.engadget.com/2222542/anthropic-says-opus-5-can-nearly-match-its-top-performing-model-for-half-the-price/
[3] https://it.slashdot.org/story/26/04/07/2115208/anthropic-unveils-claude-mythos-powerful-ai-with-major-cyber-implications
Does this end them sooner, or is it irrelevant? (Score:2)
My first reaction is that if one company is selling at half the price, that company will go bankrupt at twice the speed since every transaction was already a money-loser. Customers are absolutely the worst thing for these "businesses."
But is revenue really significant? I can see how neither reducing the price to $0 or increasing their price to 10x what it was, might not really have much of a different impact on the bottom line, compared to all the debt.
Re: (Score:2)
Customer token cost is one side of the coin.
The other side is inference cost: what's it take to run?
The Chinese models are all significantly smaller than what Opus/Mythos are, supposedly. That's a big factor in their cost efficiency, and why they're cheaper.
If they've managed to make long strides in improving the efficiency/overhead of their models, it would go just as far as dropping price.
Re: (Score:3)
Although the Chinese models are all considerably smaller, it's not obvious to me that they're as good or better. They could be, but I'm not convinced the benchmarks that currently exist are stressing the models enough to establish where models actually are any good.
Big AI's business scheme (Score:2)
1. Charge less than cost price.
2. Reduce that price due to competition.
3. Profit ?
Re: (Score:2)
You forgot: 2a - cripple the model so it refuses to do what you tell it to do while Chinese open-source models don't give a sh*t about Muh Safety.
So, no good at coding then? (Score:2)
> the company has specifically avoided training the new model on cyber-related tasks
Software development is, by definition, a cyber-related task.
Re: (Score:2)
IOW I need to now run local (or cloud) Chinese models to make sure Claude isn't designing back doors into my software.
Re: (Score:2)
I guess you're not reviewing Claude's code then? I'm pretty sure that kind of back door would be easily spotted in a code review!
Re: (Score:2)
They mean cyber-security-related tasks. If you read a bit further, it's trained on general coding, but not as much on stuff like decoding binaries and exploiting vulnerabilities. I'd expect it to be a little less effective in some areas than it could be (or Mythos is) due to this
Re: (Score:2)
I knew what they meant, my comment was sarcastic. Everybody throws around the prefix "cyber" these days and have no idea what it means.
Anthropic marketing machine executing flawlessly (Score:2)
Anthropic has followed a pattern of proclaiming that their models are so good they are dangerous. Then they come back with a new model that is supposedly "less" dangerous, but still "very powerful."
Using this strategy, they've raised the profile of Mythos to the point that they can charge double for it., through government-imposed scarcity. At the same time, they've introduced this new model that's "almost as good" but "half the price". Companies are going to come running for this "heck-of-a-deal" that is a
Re: (Score:3)
Competing on price is somehow a "scam".
You are rapidly becoming the poster boy for AI Derangement Syndrome.