Opus 5.5 high vs GPT-6.1 Sol max: AA data shows Sol's value edge is mostly latency, not intelligence
Wsz2020 · reddit · 2026-10-02
- Using Artificial Analysis data, the author unpacks the apparent value of OpenAI's GPT-6.1 Sol vs Anthropic's Opus 5.5, arguing the headline "88% cheaper at max" is misleading since Opus 5.5 shouldn't be run at max in the first place.
- Fair comparison (Opus high vs Sol max): Intelligence Index 54 vs 52, cost per task $1.82 vs $0.72, output speed 73 t/s vs 64-66 t/s, and time-to-first-token 39s vs 273s. That's $0.55 per extra index point — about $11,000/month at 10,000 tasks.
- Key insight: Sol's effort ladder is nearly flat at the top — high scores 50 at $0.32/task and xhigh 51 — so max mostly buys you latency, not intelligence. Bottom line: pick Opus 5.5 high for latency or best quality; Sol high is the cheapest near-equal option for batch workloads.
More from Models
- JevBench to add evals for LLM routing, RAG retrieval, and moderation use cases — airesearch12 · 2026-10-02
- Google's Argon battle-tested by 200k+ Googlers daily, not benchmaxxed — Zergylord · 2026-10-02
- JEV claims to be the first System One model hosted in the EU — juanviera23 · 2026-10-02
- Gemini knew a user's mom's name unprompted, then gave three conflicting explanations — Exact_Firefighter864 · 2026-10-02
- GPT-6.1 Sol nearly matches Astra at a quarter of the price on RareBench — danielmckinn0n · 2026-10-02
- New results from PostTrainBench v1.2 are in — mariofilhoml · 2026-10-02