Microsoft-Decision-1
Microsoft's hosted decision model announcement, claiming top accuracy on 36 benchmarks and low latency versus GPT-6 Sol.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
Microsoft-Decision-1 is Microsoft's hosted decision model, announced via a post on its Foundry platform. According to the listing, Microsoft claims the top accuracy across 36 benchmarks, 35 times lower median latency than GPT-6 Sol, and that the model flips its decision on only 1.3% of cases across eight perturbation styles. The post was later updated to add Jev's accuracy and calibration numbers for comparison.
No post text or model documentation was available here, so the benchmarks and methodology are not confirmed. It is relevant as a major vendor entering the typed decision model category and as a competitor benchmark against Jev.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Accuracy, latency and robustness to perturbations across 36 decision benchmarks (measured)(inferred) | — |
How it works
- Read Microsoft's announcement
- Compare reported accuracy, latency and perturbation robustness with Jev
Strengths
- Large-vendor benchmark of decision models
- Includes perturbation robustness
Caveats
- Vendor-reported claims
- Post text not available here
- Methodology not verified
At a glance
- Type
- Benchmark
- Status
- production
- Runtime
- hosted model (Microsoft Foundry)
Use cases and integrations
How curated lists describe it
Competitor benchmark: Microsoft's hosted decision model claims the top accuracy across 36 benchmarks and uses 35x lower P50 latency than GPT-6 Sol, and flips its decision on only 1.3% of eight perturbation styles, in a post later updated to add Jev's accuracy and calibration numbers.
yibie/awesome-jev
Frequently asked questions
What does Microsoft claim?
Top accuracy across 36 benchmarks, 35 times lower latency than GPT-6 Sol and 1.3% decision flips under perturbation.
Does it compare with Jev?
The post was later updated to add Jev's accuracy and calibration numbers.
Are the claims independent?
No, they are Microsoft's own.
Related projects
Japanese research deck, article and source ledger tracking TypeSafe's Jev and System One models over time.
ProjectdiscussionMarkdown- jev-arena★ 122
Side-by-side tool comparing Jev and DeepSeek on labelling the same comments for relevance, sentiment and intent, with replay and reports.
ProjectbetaTypeScript Evaluation suite for decision models with 23,900 pinned rows across nine task families and three primitives, plus a leaderboard.
ProjectbetaPythonOpen-source BYOK arena where Jev faces another model as a judge on your question, with blind voting before costs are revealed.
ProjectbetaTypeScriptpytest-style tool that pins fixtures and expected Jev answers, then fails a model upgrade when answers flip or confidence drops.
ProjectbetaPythonTest harness for any Jev-API decision model: accuracy, calibration and robustness probes for reordering, injection and padding.
ServicebetaPython