Anthropic's frontier model Claude Fable 5 continues to rank highly across multiple categories on the community-voting evaluation platform Arena.ai even after a redeployment, according to a preview of the latest tallies.
June 2026 · Anthropic · Arena.ai
Claude Fable 5 holds #1 in Code Arena — even after its re-release
Anthropic's top-tier "Mythos-class" model was paused over cybersecurity concerns, redeployed on a new endpoint, and still tops the community-voted frontend leaderboard. The re-release lost only ~27 Elo — within the confidence interval, not statistically significant.
#1
Code Arena: Frontend
80%+
Reported score on
1M
Default context
Code Arena: Frontend — top of the leaderboard
Elo score (higher = better). GLM-5.2 has closed the gap, tightening the frontier race.
The cost of the frontier: Fable 5 vs Opus 4.8
Output pricing is roughly double, and token consumption can run 3× higher.
OUTPUT PRICE ($ / M tokens)
TOKEN CONSUMPTION (relative)
WHAT DEVELOPERS PRAISE
Fixed a legacy Shopify auth issue in one shot where Opus 4.8 and GPT-5.5 failed
Stable over long sessions: 3D model generation, Factorio automation, CAD editor builds
Tuned for long-horizon agentic work with self-verification and long-term memory
THE DRAWBACKS
Token use over 3× Opus 4.8; high price and fast rate-limit exhaustion
Safety classifiers trigger easily, routing even harmless prompts to Opus 4.8
Seen by some as overkill and cost-inefficient for everyday tasks
Paused after launch (~June 12) over a U.S. government directive and cybersecurity concerns, then redeployed (~July 1) with updated safeguards.
This Arena.ai comparison counts only votes on the new endpoint — where Fable 5 still leads across Text, Agent, Design and Frontend, keeping the accumulation of future votes a point to watch.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…