MissionReport benchmark: frontier models score just ~60% on government status reporting
Exp_Mark · x · 2026-10-27
micro1.ai released MissionReport, a benchmark measuring frontier models' judgment on government status reporting. All 20 tasks were written by experts who have done the job (weekly activity reports, program status reports, after-action reviews); models must produce a finished Word document graded by expert rubrics. Nine frontier models ran each task three times, yielding 540 graded reports.
Key findings:
- Models struggle: scores range from 54.3% (Gemini 3.8 Flash) to 64.1% (Gemini 4 Argon), and only 3 of 540 reports met every rubric criterion.
- Deciding what to include and synthesizing the point a senior reader needs account for 67–77% of every model's lost points, while models earn 80–91% on formatting.
- Models are especially weak at flagging unresolved issues and separating evidence-backed conclusions from noise.
The takeaway: as government staff increasingly draft reports with AI, models remain far from reliable on the core judgment that makes a report useful.
More from Models
- Week 3 of "Fable 5.5 launches this week" hype circles: no new frontier models this October — daniel_mac8 · 2026-10-27
- Speculative leaderboard shows "Gemini 4 Argon" vs "GPT-6.1 Sol"; observer calls MiniMax parity suspicious — teortaxesTex · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27
- Kimi K3.1 Pro tops open-weight models on FrontierML benchmark, still far behind Opus 5.5 — teortaxesTex · 2026-10-27
- Across 7 tasks, highest reasoning setting isn't the best for most open models — keunwoochoi · 2026-10-27
- Open Models Match GPT-6 on MedMCQA but Trail 17 Points on Hard Clinical Cases — keunwoochoi · 2026-10-27