Opus 5.5 benchmark rundown ends with hopes for a GPT-6 Sol matchup
The post lists Opus 5.5's benchmark scores, pricing and token usage, then says it looks forward to GPT-6 Sol competing head to head.
Anthropic just released Opus 5.5. A brutal successor to Opus 5, and apparently a massive improvement over the Claude family. The benchmarks are pretty insane too. Here’s what we have: ⟢ Mogs Astra & Fable 5.1 on Terminal-Bench. ⟢ Beats Fable 5.1 on AA Intelligence Index, scoring 57.6 against 53.4. ⟢ Performs at Fable 5.1’s level on most tasks, while generating output over 30% faster than Opus 5. ⟢ Writes more naturally and follows writing instructions better. Plus there’s been a decent price cut. Opus 5.5 costs $4/$20 per million input/output tokens, 20% cheaper than Opus 5. They even claim it costs 40% less to run on typical workloads. But I like to think that’s not good enough considering it’s still a token monger. According to AA, Opus 5.5 consumed around 119K output tokens per task at maximum effort. The highest in their comparison. Which is just ridiculous. Overall I’m excited for this, OpenAI has a lot to work on with GPT-6 Sol, and I expect it to go head to head with this . Anthropic already has Sonnet 5.5 and Haiku 5.5 on the way, plus Grok 5 and the rumored Gemini 4 Pro still to come. It’s raining new models atp.
This post quotes another post: “Anthropic releases Claude Opus 5.5 model”