Anthropic · June 9, 2026
The new #1 model is here — but it costs double. The whole decision is one trade: 80.3% SWE-Bench Pro for $10/$50, or 69.2%for half the price. Here's when the frontier is worth it.
SWE-Bench Pro80.3%GDPval-AA1,932API pricing$10 / $50Context1M tokens
SWE-Bench Pro69.2%GDPval-AA1,890API pricing$5 / $25Context1M tokens
An 11-point SWE-Bench Pro jump in a single release.
Opus 4.7 → 4.8 moved the needle 4.9 points. Fable 5 more than doubles that in one step — from 69.2% to 80.3%. It's the largest one-release coding gain any frontier lab has posted in 2026, and it's why Mythos rattled Wall Street two months before this public launch.
From official Anthropic launch data · June 9, 2026
Capability went up. So did the bill.
Fable 5
Opus 4.8
Free until June 22, 2026. Anthropic dropped Fable 5's paywall at launch — test it on your real workload before the meter starts.
Fable 5 and Mythos 5 are the same model. Identical weights, identical benchmarks. The only difference is safety and access: Fable 5 ships with safeguards that block specific high-risk cyber responses and is available to everyone. Mythos 5 lifts those safeguards for advanced security research and is restricted to vetted partners. For 99% of people, Fable 5 isMythos 5 — just the version you're allowed to use.
Full Mythos 5 vs Fable 5 breakdownOne line — same API as Opus 4.8.
Identical interface, including parallel subagents. Anthropic API docs →
Related