We provide free, open source software to compute the impact of public policy. See our global account @ThePolicyEngine and UK account @PolicyEngineUK.
United States
- Tweets 557
- Following 95
- Followers 924
- Likes 321
PolicyBench has a new #1: GPT-6 Sol gets 94.2% of tax and benefit answers within $1 with no tools, ending GPT-5.6 Sol's run at the top since July. Among other models release yesterday, Claude Opus 5.5 enters second at 92.9%, GPT-6 Luna fourth at 91.4%.
policybench.org/notes/2026-0…
1
2
142
This morning, the Census announced that the SPM child poverty rate was 13.4% in 2025, down 0.1 pp from 2024.
Today we're launching the Child Poverty Impact Dashboard: a tool to analyze how reforms can affect this rate across all 50 states and DC.
policyengine.org/us/child-po…
1
2
5
593
Configure a reform: a state Child Tax Credit, EITC, dependent
exemption, child allowance, and more to see its impact on child poverty, individual households, income distribution, and congressional districts.
1
1
67
See it live in less than 90 minutes with guest speakers @BlahaOsborne
(Prenatal-to-3 Policy Impact Center) and @JoshuaTMcCabe (Niskanen
Center).
Register: us06web.zoom.us/meeting/regi…
1
121
BLS re-estimates SPM poverty thresholds every year from consumer spending, not inflation. Renter threshold for two adults and two children: +6.3% from 2024 to 2025, more than twice CPI. PolicyEngine now builds and projects those thresholds the way BLS does, from the Consumer Expenditure microdata with BLS's procedure. A calculator and a paper:
2
2
446
Everything is open: pip install spm-calculator, the audit JSON with every input hash, and the paper's replication code.
1
28
Census releases the 2025 poverty numbers today at 10am ET. At 2pm ET we go through them live with guests @BlahaOsborne and @JoshuaTMcCabe and demo the tools: "The new child poverty numbers, and how state reforms would affect them." Register: us06web.zoom.us/meeting/regi…
86
PolicyBench tests how accurately AI models compute US taxes and benefits. GPT-6 Astra debuts second of 39 at 88.0% of answers within $1, 1.2 points behind GPT-5.6 Sol.
We've also added new Gemini, DeepSeek, and GLM models.
2
2
15
2,427
Most of the gap between GPT-6 Astra and GPT-5.6 Sol is two rules Astra invented regarding payroll taxes and Medicare eligibility for people with disabilities. policybench.org/notes/2026-0…
1
4
145
New poverty numbers land the morning of Tuesday, September 15. That afternoon at 2pm ET, join us for guest remarks on the release and a live demo of PolicyEngine's new Child Poverty Impact Dashboard.
Register: us06web.zoom.us/meeting/regi…
1
3
4
575
What would a state Child Tax Credit, an expanded EITC, or a child allowance do to child poverty in your state? The dashboard answers in minutes, for all 50 states and DC, down to the congressional district.
1
1
64
Guest speakers: @BlahaOsborne of @pn3policy at Vanderbilt, and @JoshuaTMcCabe of the @NiskanenCenter. Then @DTrim_99 and @MaxGhenis walk through the tool live, with Q&A to close.
us06web.zoom.us/meeting/regi…
2
262
PolicyBench tests how accurately AI models compute US taxes and benefits. Two new rows: Ox Alpha, a cloaked preview model with no maker named, debuts fourth of 32 at 84.2% of answers within $1. @grok 4.6 debuts eighth at 82.8%, 1.9 points past Grok 4.5.
policybench.org
1
1
6
1,768
PolicyBench tests how accurately AI models compute US taxes and benefits. New: Gemini 3.7 Flash scores 79.2% of answers within $1 — tenth of 30, just past its predecessor, at $0.014 per household.
Full board: policybench.org
121
PolicyBench, our benchmark of AI accuracy on US taxes and benefits, found forcing a tool call silently turns off Claude's thinking — others reason regardless. Under tool_choice auto, Claude Fable 5 jumps 79.9→86.9% within $1: would-be #2.
More: github.com/PolicyEngine/poli…
1
10
1,452