Skip to content

Sam's News — ai — Week 2026-W30

TL;DR

  • Anthropic launched Claude Opus 5, matching or beating rival frontier models on key benchmarks at roughly half the cost, signaling an industry pivot toward cost-efficiency.

Biggest stories

Anthropic's Claude Opus 5 launch

Anthropic released Claude Opus 5 on July 25, priced at $5/$25 per million input/output tokens — unchanged from Opus 4.8 and half rival Fable 5's $10/$50. It scored 43.3% on Frontier-Bench v0.1 (vs. Opus 4.8's 18.7%, Fable 5's 33.7%), more than doubling Opus 4.8's prior software-engineering score, and came within 0.5% of Fable 5 on CursorBench 3.2 at half the cost. It topped the Artificial Analysis Index at 61 points, scored 3x higher than the next-best rival on ARC-AGI 3, and beat Fable 5 on OSWorld 2.0 at one-third the cost. Features include a new "effort dial" (low/medium/high), a fast mode at ~2.5x speed for double price, a 1M-token context window, and a May 2026 knowledge cutoff. Trained without offensive cyber capabilities, cutting safety interventions ~85% versus Fable 5. Now the default model on Claude Max and the strongest available on Claude Pro.

Sources: 2026-07-25, 2026-07-26

  • Single dominant storyline for the week: Anthropic's Opus 5 launch, covered two days running as benchmark details and third-party analyses emerged.
  • Cost-efficiency, not raw capability, framed as the key competitive axis against rival model Fable 5.