Claude Fable 5 and Mythos 5 Launch with New Pricing

Anthropic just cut the price of their high-end models by more than half. Fable 5 and Mythos 5 are now $10 per million input tokens and $50 per million output tokens. It's a massive drop from the Mythos Preview pricing, and it's a move that tells us exactly where the industry is heading.

We're seeing a pattern where "state-of-the-art" is becoming a commodity. Anthropic claims Fable 5 beats everything they've ever released in software engineering and scientific research. That might be true, but the real story isn't the benchmark scores. It's the fact that high-end intelligence is getting cheap enough to actually use at scale without bankrupting a project.

I'm skeptical about how long these margins can last, but for now, the barrier to entry for heavy-duty LLM workloads just collapsed. The question is whether the performance jump in Fable 5 actually justifies the switch, or if we're just chasing numbers on a spreadsheet.

The new pricing structure

The pricing is $10 per 1 million input tokens and $50 per 1 million output tokens. Compared to the Mythos Preview rates, this is a price hike. It's a gamble on the model's capability, but for most developers, the cost is just the cost of doing business if the model actually works.

The model is a step forward in handling "long-horizon" problems. This is a fancy way of saying it doesn't lose the plot when you ask it to refactor a system across five different files. In early testing, it showed more autonomy and reliability than previous versions on CursorBench. Whether that reliability justifies the $50 output cost depends on how many tokens you're burning on boilerplate.

If you're automating these calls, you'll want to track your spend closely so you don't wake up to a surprise bill.

def calculate_cost(input_tokens, output_tokens):
    input_cost = (input_tokens / 1_000_000) * 10
    output_cost = (output_tokens / 1_000_000) * 50
    return input_cost + output_cost

print(f"Total cost: ${calculate_cost(10000, 2000):.4f}")

Accessibility and scale

The price drop here is the only part that actually moves the needle. Cutting the cost of input and output tokens by more than half compared to the Mythos Preview isn't just a marginal update; it's a signal that the cost of intelligence is crashing faster than the marketing teams want to admit. For a lot of devs, this moves these models from "expensive experiment" to "actually viable for production." I think the claim about "bringing capabilities to as many users as possible" is just corporate fluff, but the numbers themselves are real.

As for the panic over bug bounty hunters using these models to find more vulnerabilities—I think that's a distraction. Security researchers have always used the best tools available to break things. If a $10-per-million-token model makes it easier to find a memory leak or a logic flaw, that's a problem with the code, not the tool. I'd argue that the real risk isn't the "sophisticated" reports, but the sheer volume of low-effort noise that will now flood security teams' inboxes.

I'm still not convinced this solves the latency issue for real-time applications. We have cheaper tokens, sure, but we don't have faster ones. The question is whether we've hit a wall where cost no longer matters because the bottleneck has shifted entirely to inference speed.

Conclusion

The new pricing tiers make these models accessible, but they don't magically solve the cost problem for anyone running massive workloads. You're still paying for the luxury of a larger context window, and the gap between Fable and Mythos is mostly a question of how much you're willing to bleed for a slight bump in reasoning.

I'm still not convinced that the pricing for Mythos is justified for the average dev. It feels like a premium tax on a marginal gain.

The real question is: at what point does the cost of these high-end tokens stop being a line item and start becoming a liability for your margins?