Reported by others
Operating a computer screen (OSWorld 2.0)
The model looks at a real operating system screen and uses the mouse and keyboard to finish tasks that span several apps.
How to read this table
- Higher rows are rated higher. A range like “1–2” means the gap is within the margin of error.
- The score is the value published by the task's authors and other organizations.
- Conditions can differ between models, so small gaps do not mean much.
- Prices are a guide per million tokens, reading then writing.
- Length is how much text the model can read at once (tokens).
| Rank | Model | Developer | Success rate (%) | Input $ /1M | Output $ /1M | Context | Supports | Weights | Released | Price and specs |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5 | Anthropic | Rating 31.4 | $5 | $25 | 1M | Anthropic Released 2026-07 Price $5 / $25 Length 1M | |||
| 2 | GPT-5.6 Sol | OpenAI | Rating 27.3 | $4 | $20 | 1.05M | OpenAI Released 2026-07 Price $4 / $20 Length 1.05M | |||
| 3 | Claude Opus 4.8 | Anthropic | Rating 20.6 | $5 | $25 | 1M | Anthropic Released 2026-05 Price $5 / $25 Length 1M | |||
| 4 | Claude Opus 4.7 | Anthropic | Rating 18.2 | $5 | $25 | 1M | Anthropic Released 2026-04 Price $5 / $25 Length 1M | |||
| 5 | GPT-5.5 | OpenAI | Rating 13.0 | $5 | $30 | 1.05M | OpenAI Released 2026-04 Price $5 / $30 Length 1.05M | |||
| 6 | Claude Sonnet 4.6 | Anthropic | Rating 9.3 | $3 | $15 | 1M | Anthropic Released 2026-02 Price $3 / $15 Length 1M | |||
| 7–8 | Kimi K2.6 | Moonshot | Rating 4.6 | $0.95 | $4 | 262.1K | Moonshot Released 2026-04 Price $0.95 / $4 Length 262.1K | |||
| 7–8 | MiniMax-M3 | MiniMax | Rating 4.6 | $0.3 | $1.2 | 1.05M | MiniMax Released 2026-06 Price $0.3 / $1.2 Length 1.05M | |||
| 9 | Qwen3.7 Plus | Alibaba | Rating 2.8 | $0.4 | $1.6 | 1M | Alibaba Released 2026-06 Price $0.4 / $1.6 Length 1M |
Best success rate over time
2026-02-17Release date2026-07-24
How to read this page
- The value is the share of tasks finished. Figures published by the organizations that made the tasks, among others, collected here.
- The success rate is the share of tries that succeeded. A gap of 1 point is 1 success out of 100 tries.
- In the Supports column the icons are, from left: can use tools / can read images / can answer in a fixed format (JSON); faded means not supported or unknown.
- Open-weights models have their contents (the weights) published, so you can also run them on your own computer or server. Models without the open-weights mark are normally used through the developer's own services (apps or API); for some models the weights status is unknown.
- Names such as ChatGPT next to a model are the consumer services its developer offers. Check on the service itself whether that model is available there.
Data: Arena, Epoch AI, Swallow LLM team (CC BY 4.0), Models.dev (MIT), updated 2026-10-24, data sources