Breaking

GPT-6 Astra takes Code Arena WebDev lead over Claude Fable 5.1

September 5, 2026 at 19:15 EDT

Foundation Models
Software Dev & Coding

OpenAI’s GPT-6 Astra (Max) has moved to first place on Arena.ai’s Code Arena WebDev leaderboard with 1,797 points, 35 points ahead of Anthropic’s Claude Fable 5.1 (Max) at 1,762, two days after Astra’s September 3, 2026 launch. Claude Opus 5 (Max) sits third at 1,688. Arena.ai described the result as a new high-water mark on the board and as a shift in the cost-performance frontier at $40 per million tokens, matching recent Claude pricing.

September 5, 2026 · Code Arena WebDev

Astra takes the coding crown

GPT-6 Astra (Max) is first on Arena.ai’s community WebDev board at 1,797 points, 35 points ahead of Claude Fable 5.1 (Max). Claude Opus 5 (Max) is third at 1,688.

1797

Astra (Max) score

+35

pts over Fable 5.1

1688

Opus 5 (Max) third

+180

vs GPT-5.6 Sol

WebDev Elo, drawn to scale above 1,650

Column height = score minus 1,650, so the 35-point and 109-point gaps are visible.

1797

Astra

1762

Fable 5.1

1688

Opus 5

What testers like

Astra is now treated as Fable-class on many agentic frontend jobs, with early category leads in data, consumer product, and content tools.

What remains contested

Independent coding indices still put Fable 5.1 slightly ahead (70 vs 67 on one agent index). Some larger app builds still favor Fable, and several Code Arena categories are unfinished.

Same list price, different cache math

Both flagships list $10 / $50 per million input/output tokens. Fable 5.1 cache reads are $0.25/MTok; Astra cached input is $1.00/MTok. Arena called Astra the best result at a $40/MTok blended point on this board.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.