Breaking

Claude Fable 5.1 Outscores GPT-5.6 Sol on Agentic Benches as Token Prices Diverge

September 5, 2026 at 17:35 EDT

Foundation Models
AI Agents
Software Dev & Coding

Anthropic's Claude Fable 5.1, launched September 1, 2026, leads OpenAI's GPT-5.6 Sol on Anthropic's published agentic science and coding evaluations, while Sol remains far cheaper on current promotional API rates.

Sep 1, 2026 · Anthropic / OpenAI

Fable 5.1 takes on GPT-5.6 Sol

Anthropic's new Mythos-class model leads OpenAI's flagship on published agentic science and coding benches. Sol remains far cheaper on promotional token rates.

52.6%

Fable 5.1 on Terminal-Bench-Science

22.4%

GPT-5.6 Sol on the same science bench

2.5×

Sol cheaper on current promo sticker

66 vs 61

Intelligence Index, Fable 5.1 vs Sol (max)

Agentic science, drawn to scale

Terminal-Bench-Science 0.1 accuracy. Column height is proportional to the score.

22.4%

GPT-5.6 Sol

24.7%

Fable 5

52.6%

Fable 5.1

Fable 5.1 more than doubles both its predecessor and Sol on this vendor-run science harness.

Sticker price, same unit

Input dollars per million tokens. Each block is $2.

GPT-5.6 Sol
promo in

$10

Fable 5.1
list in

Takeaway: Sol is 2.5× cheaper on input (and on output) at the current promo. Fable 5.1 instead cuts cache reads to $0.25.

Where Fable 5.1 is winning

Hard long-horizon jobs, vendor science and terminal benches, and early partner reports of unattended runs, crash diagnosis, and more natural writing.

Where Sol still wins on paper

Promo token rates through at least Nov 21, 2026. Agentic rows are vendor-run; at max effort Fable 5.1 can spend more output tokens, and long Sol prompts surcharge the whole request.

Continue reading

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