Today, OpenAI released GPT-6 Astra. The company calls it its most intelligent and aligned model, and positions it primarily as a computer-use system rather than a chat model. The pitch is that Astra operates software the way a person does, across browsers, spreadsheets, desktop applications and terminals, and finishes multi-step jobs instead of describing how to do them.

Is it deployable? Partly, and not on your own hardware. Astra is a closed, hosted model with no released weights, so self-hosting is not an option. It is live today only for organizations in OpenAI’s Trusted Access and Daybreak programs.

What is actually new

The main change for devs is context handling. Codex previously used compaction, summarizing earlier turns once context filled up. That process discards the detail an agent later needs: why a fix failed, which tests ran, which requirement was added early. Astra instead keeps notes across context windows and searches back into earlier messages and tool output. The feature ships experimental behind a config.toml setting and becomes the Codex default in the coming weeks.

Astra can also ask the user a question while continuing work that does not depend on the answer. That removes a common agent failure where one unresolved decision stalls an entire job.

On the model page, Astra lists a 1,050,000-token context window, 128,000 max output tokens and an April 30, 2026 knowledge cutoff. Input is text and image, output is text only. reasoning.effort adds two new levels above high: xhigh and max. Tool support covers computer use, hosted shell, apply patch, skills, MCP and tool search. Fine-tuning is not supported.

The benchmark picture

OpenAI reports 72.6% on OSWorld V2-Offline against 65.7% for GPT-5.6 Sol, with average task time falling from roughly 75 minutes to 40. Anthropic reports 77.9% for Claude Fable 5.1 but says it used a different OSWorld release and should not be compared directly.

Astra scores 98.6% on ARC-AGI-3. That number was produced with a Responses API harness that retains reasoning between turns and uses compaction for long contexts, and OpenAI has previously shown those settings move ARC-AGI-3 scores substantially without changing the model. The result measures the model plus the agent system.

Other reported figures: 97.6% on FrontierMath Tier 4, 95.9% on BenchCAD Vision2Code against 84.3% for Fable 5.1, and 64.6% on Terminal-Bench Science against Anthropic’s reported 52.6%. Epoch AI notes OpenAI funded FrontierMath and has exclusive access to part of it.

Coding is the weak spot in the story. Astra scores 74.1% on DeepSWE v1.1 versus 70.8% for Sol. Meta reported 75.4% for Muse Spark 1.3 at maximum reasoning, and the public leaderboard puts Gemini 3.8 Flash and Claude Opus 5 near 74%. On a 113-task benchmark, those gaps are one or two tasks.

‘;wrap.insertAdjacentHTML(‘beforeend’,h);
});
function run2(){var f=document.querySelectorAll(‘#bars .fill’);
f.forEach(function(x){x.style.width=”0″});
setTimeout(function(){f.forEach(function(x,i){setTimeout(function(){x.style.width=x.dataset.w+’%’},i*70)})},60);}

// panel 3
var inT=document.getElementById(‘inTok’),outT=document.getElementById(‘outTok’);
function fmt(n){return n.toLocaleString(‘en-US’)}
function cost(i,o,pi,po){var m=i>272000?1:0;var ci=(i/1e6)*pi*(m?2:1);var co=(o/1e6)*po*(m?1.5:1);return ci+co;}
function calc(){var i=+inT.value,o=+outT.value;
document.getElementById(‘inLab’).textContent=fmt(i);
document.getElementById(‘outLab’).textContent=fmt(o);
var a=cost(i,o,10,50),s=cost(i,o,5,30);
document.getElementById(‘cAstra’).textContent=”$”+a.toFixed(2);
document.getElementById(‘cSol’).textContent=”$”+s.toFixed(2);
document.getElementById(‘cRatio’).textContent=(a/s).toFixed(1)+’x’;}
inT.addEventListener(‘input’,calc);outT.addEventListener(‘input’,calc);calc();

run0();run2();
window.addEventListener(‘load’,sync);setTimeout(sync,200);setTimeout(sync,900);
})();

“>