MyPCBench evaluates assistants on 184 tasks across a personal desktop with 17 simulated web apps. Tasks involve personal records, travel, shopping, calendars, and coordinating work across applications. These results use the canonical computer-and-bash setup with a 100-step limit and report the percentage of tasks where every rubric item passes.
| Model / creator | Version | Reported score |
|---|---|---|
 Claude Opus 4.6Anthropic · Proprietary | 184 tasks · 100 steps | 58.2% |
 GPT-5.6 LunaOpenAI · Proprietary | 184 tasks · 100 steps | 55.4% |
| GPT-5.6 SolOpenAI · Proprietary | 184 tasks · 100 steps | 55.4% |
| Claude Sonnet 4.6Anthropic · Proprietary | 184 tasks · 100 steps | 50.5% |
| GPT-5.5OpenAI · Proprietary | 184 tasks · 100 steps | 45.1% |
| GPT-5.4 miniOpenAI · Proprietary | 184 tasks · 100 steps | 23.9% |
 Qwen 3.5 35B-A3BQwen · Open source / open weights | 184 tasks · 100 steps | 7.6% |
| Qwen 3.5 9BQwen · Open source / open weights | 184 tasks · 100 steps | 6.5% |
interface: computer + bash · judge: gemini-3.1-flash-lite-preview · max steps: 100 · rubric score percent: 85.1 · selection: Canonical matched-protocol rows only; four 200-step self-reported rows retained in source snapshot, excluded from this track. · tasks: 184
MyPCBench / Carnegie Mellon University ↗interface: computer + bash · judge: gemini-3.1-flash-lite-preview · max steps: 100 · rubric score percent: 83.2 · selection: Canonical matched-protocol rows only; four 200-step self-reported rows retained in source snapshot, excluded from this track. · tasks: 184
MyPCBench / Carnegie Mellon University ↗interface: computer + bash · judge: gemini-3.1-flash-lite-preview · max steps: 100 · rubric score percent: 85.3 · selection: Canonical matched-protocol rows only; four 200-step self-reported rows retained in source snapshot, excluded from this track. · tasks: 184
MyPCBench / Carnegie Mellon University ↗interface: computer + bash · judge: gemini-3.1-flash-lite-preview · max steps: 100 · rubric score percent: 78.5 · selection: Canonical matched-protocol rows only; four 200-step self-reported rows retained in source snapshot, excluded from this track. · tasks: 184
MyPCBench / Carnegie Mellon University ↗interface: computer + bash · judge: gemini-3.1-flash-lite-preview · max steps: 100 · rubric score percent: 83.6 · selection: Canonical matched-protocol rows only; four 200-step self-reported rows retained in source snapshot, excluded from this track. · tasks: 184
MyPCBench / Carnegie Mellon University ↗interface: computer + bash · judge: gemini-3.1-flash-lite-preview · max steps: 100 · rubric score percent: 53.5 · selection: Canonical matched-protocol rows only; four 200-step self-reported rows retained in source snapshot, excluded from this track. · tasks: 184
MyPCBench / Carnegie Mellon University ↗interface: computer + bash · judge: gemini-3.1-flash-lite-preview · max steps: 100 · rubric score percent: 42.4 · selection: Canonical matched-protocol rows only; four 200-step self-reported rows retained in source snapshot, excluded from this track. · tasks: 184
MyPCBench / Carnegie Mellon University ↗interface: computer + bash · judge: gemini-3.1-flash-lite-preview · max steps: 100 · rubric score percent: 24.1 · selection: Canonical matched-protocol rows only; four 200-step self-reported rows retained in source snapshot, excluded from this track. · tasks: 184
MyPCBench / Carnegie Mellon University ↗Humanity’s Last Exam evaluates advanced knowledge and reasoning across a broad range of academic subjects using expert-written questions.
LiveBench: ReasoningLiveBench 2.0 evaluates reasoning with LiveBench on its October 25, 2026 release.
AutomationBench-AAAutomationBench-AA evaluates agents working across simulated business applications.
