Ranked by OSWorld - 21 models
Best AI models for computer use
In board data through July 31, 2026, Claude Fable 5 ranks first on the official OSWorld (OSWorld-Verified) leaderboard with 86.0%, 2.6 points ahead of Claude Opus 5 at 83.4%. Its list price is $10 input / $50 output per 1M tokens. A fixed rule, the lowest output price per 1M tokens among models with a registry price that score within 5 points of the top score, picks Claude Opus 5 at 83.4% ($5 input / $25 output per 1M tokens). The highest-placed open-weight model is MiniMax-M3, ranked 5th with 75.2%.
Board data through , checked . How this is ranked
OSWorld (OSWorld-Verified) ranking
| Rank | Model | Provider | OSWorld | Source | Input / 1M | Output / 1M | Context | Weights |
|---|---|---|---|---|---|---|---|---|
| 1 | Claude Fable 5Board: claude-fable-5[1m] (Anthropic) | Anthropic | 86.0%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 307.73 / 358; rank shown as 1 with an asterisk on the leaderboard | osworld-v1.xlang.ai | $10 | $50 | 1M | Closed |
| 2 | Claude Opus 5Board: claude-opus-5[1m] (Anthropic) | Anthropic | 83.4%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 300.2 / 360 | osworld-v1.xlang.ai | $5 | $25 | 1M | Closed |
| 3 | Muse Spark 1.1 (Meta Superintelligence Labs) | - | 80.7%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 291.22 / 361 | osworld-v1.xlang.ai | - | - | - | - |
| 4 | Holo3-35B-A3B (H Company) | - | 80.4±2.2%Official leaderboard - OSWorld-Verified Results, Type: Specialized model, Max Steps: 100, Runs: 2 (Run 1: 296.41/359, Run 2: 280.55/359) | osworld-v1.xlang.ai | - | - | - | - |
| 5 | MiniMax-M3Board: MiniMax M3 (MiniMax) | MiniMax | 75.2%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 269.18 / 358 | osworld-v1.xlang.ai | $0.30 | $1.20 | 1M | Open |
| 6 | Qwen 3.7 Plus (Qwen Team, Alibaba Group) | - | 73.3%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1 | osworld-v1.xlang.ai | - | - | - | - |
| 7 | Kimi K2.6Board: Kimi K2.6 (Moonshot AI) | Moonshot AI | 73.1%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 261.58/358 | osworld-v1.xlang.ai | $0.95 | $4 | 262K | Open |
| 8 | Claude Sonnet 4.6Board: claude-sonnet-4-6 (Anthropic) | Anthropic | 72.1%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 256.71/356 | osworld-v1.xlang.ai | $3 | $15 | 1M | Closed |
| 9 | Kimi K2.5 (Moonshot AI) | - | 63.3%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 228.65/361 | osworld-v1.xlang.ai | - | - | - | - |
| 10 | Claude Sonnet 4.5Board: claude-sonnet-4-5-20250929 (Anthropic)+2 more rows on the board | Anthropic | 62.9%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 226.35/360 | osworld-v1.xlang.ai | $3 | $15 | 200K | Closed |
| 11 | Seed-1.8 (ByteDance Seed) | - | 61.9%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 222.12/359 | osworld-v1.xlang.ai | - | - | - | - |
| 12 | EvoCUA-20260105 (Meituan LongCat Team) | - | 56.7%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 50, Runs: 1, Success/Total 204.23/360 | osworld-v1.xlang.ai | - | - | - | - |
| 13 | GUI-Owl-1.5 32B (Tongyi Lab, Alibaba Group) | - | 55.4%Official leaderboard - OSWorld-Verified Results, Type: Specialized model, Max Steps: 50, Runs: 1, Success/Total 197.37/356 | osworld-v1.xlang.ai | - | - | - | - |
| 14 | DeepMiner-Mano-72B (Mininglamp Technology) | - | 53.9%Official leaderboard - OSWorld-Verified Results, Type: Specialized model, Max Steps: 100, Runs: 1, Success/Total 194.62/361 | osworld-v1.xlang.ai | - | - | - | - |
| 15 | UI-TARS-2-2509 (ByteDance Seed) | - | 53.1%Official leaderboard - OSWorld-Verified Results, Type: Specialized model, Max Steps: 100, Runs: 1, Success/Total 191.69/361 | osworld-v1.xlang.ai | - | - | - | - |
| 16 | EvoCUA (Meituan LongCat Team) | - | 50.3%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 50, Runs: 1, Success/Total 181.06/360 | osworld-v1.xlang.ai | - | - | - | - |
| 17 | EvoCUA-8B-20260105 (Meituan LongCat Team) | - | 46.1%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 50, Runs: 1, Success/Total 166.28/361 | osworld-v1.xlang.ai | - | - | - | - |
| 18 | opencua-72b-preview (The University of Hong Kong & Moonshot AI)+1 more row on the board | - | 45.0±0.9%Official leaderboard - OSWorld-Verified Results, Type: Specialized model, Max Steps: 100, Runs: 3 (Run 1: 165.89/360, Run 2: 157.40/359, Run 3: 161.73/359) | osworld-v1.xlang.ai | - | - | - | - |
| 19 | claude-4-sonnet-20250514 (Anthropic)+1 more row on the board | - | 43.9%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 50, Runs: 1, Success/Total 158.42/361 | osworld-v1.xlang.ai | - | - | - | - |
| 20 | UI-TARS-250705 (ByteDance Seed) | - | 41.8%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 151.06/361 | osworld-v1.xlang.ai | - | - | - | - |
| 21 | qwen3-vl-flash-2025-10-25 (Qwen Team, Alibaba Group) | - | 41.6%Official leaderboard - OSWorld-Verified Results, Type: General model, Max Steps: 100, Runs: 1, Success/Total 150.07/361 | osworld-v1.xlang.ai | - | - | - | - |
Prices in USD per 1M tokens at the standard rate, from each provider's pricing page. A dash means the board name does not match a model in our registry.
Related
API capabilities
Providers of the top 3
Curated picks
FAQ
What is the best AI model for computer use?
Claude Fable 5 ranks first by OSWorld (OSWorld-Verified), board data through July 31, 2026, with 86.0% (official leaderboard). One benchmark is not a verdict: settings differ between rows, so test on your own workload before you switch.
What is the cheapest AI model for computer use near the top of OSWorld?
Taking every ranked model within 5 points of the top score that has a price in our registry, Claude Opus 5 has the lowest output price: $5 input / $25 output per 1M tokens, with 83.4% by OSWorld (OSWorld-Verified), board data through July 31, 2026.
What is the best open-weight model for computer use?
MiniMax-M3 is the highest-placed open-weight model by OSWorld (OSWorld-Verified), board data through July 31, 2026, ranked 5th with 75.2%.
How is this computer use ranking built?
Rows follow the OSWorld (OSWorld-Verified) board's own order, checked on October 17, 2026; its newest entry is dated July 31, 2026, and a use case is only published while that date is within 120 days of the check. 4 board rows repeat a model already listed under a different agent, effort level or step budget; each model keeps its first, highest-placed row, in the board's own order. Prices, context windows and weights come from each provider's first-party pricing page in our model registry; a model the registry does not list shows a dash.
Newsletter
Get new benchmark results as they land
One email per week. New model releases, the leaderboard moves that matter and what they mean for what you build.