OSWorld V2
AssessTools
A benchmark for AI computer use and cross-application task execution.
Why it's here
Placed in Assess: 2 article(s) of evidence from 2 source(s), led by model releases, with 2 in the last 30 days. Confidence 37%.
Evidence (2)
- 7The New Stack·10/19/2026model_releaseMiniMax halves M3.1 Flash token prices, but agent savings remain uncertain
MiniMax launched M3.1 Flash with per-token prices cut in half and claimed up to 70% lower agent costs, but that figure depends on workload and token usage that the company did not fully disclose. Independent comparisons show lower average task costs than its predecessor, while benchmark results suggest competitive coding performance but weaker computer-use results and a relatively high hallucination rate on factual questions.
- 7Hacker News·10/19/2026model_releaseMiniMax releases M3.1 Flash for agentic tasks
MiniMax has officially launched M3.1 Flash, a next-generation model focused on agentic workflows, coding, computer use, and cybersecurity. The company says it is faster and cheaper than its predecessor, with online inference nearly 3x faster and usage costs reduced by 70%.