CMU's cua-speedrun ranks computer-use agents on speed and cost: GPT-6.1 Terra fastest, Opus 5.5 top score 85.1%
kohjingyu · x · 2026-11-04
CMU researchers (Aggarwal, Welleck, Salakhutdinov et al.) release cua-speedrun, a computer-use agent benchmark that measures not just success but task time and model-call cost. All runs use the same pipeline and pay-per-second Modal VMs through one agent interface across OSWorld, OSWorld 2.0, CUA-World, and MyPCBench. Early results: gpt-6.1-terra is fastest and cheapest (487.6s, $0.362/task at 54.9%); Opus 5.5 holds the top score at 85.1% (3050s, $15.17/task); Meta's Muse Wonder ranks third behind only OpenAI and Anthropic models.
More from Models
- '6 Astra High' spotted in model picker, hinting at new Astra release — gethackteam · 2026-11-04
- Blogger: Everyone's Sleeping on Google's Latest Models — 'They're Back' — iruletheworldmo · 2026-11-04
- Blogger Claims OpenAI Astra 6.2 Cuts Price 60% to Match Opus 5.5 — bindureddy · 2026-11-04
- Sarvam Prism launch demoed with a Snake game comparing decision models vs LLMs — itsOmSarraf_ · 2026-11-04
- Aohan: inference should be fast enough that the environment becomes the bottleneck — _arohan_ · 2026-11-04
- Cohere adds open models to Model Vault, with Encrypted tier hiding data even from Cohere — cohere · 2026-11-04