xAI

grok-4.7

Made by xAI. Released 2026-09-21. Accessibility not recorded. Cheapest standard rate we hold: $6 per million output tokens, at < 200k prompt tokens context.

What it costs

TierKindInOutCached inFrom
standardtext · < 200k prompt tokens$2$6$0.50seller
standardtext · ≥ 200k prompt tokens$4$12$1seller

per million tokens, US dollars. * marks a rate from a single registry or two that disagree — the best available number rather than a confirmed one. Every other row is confirmed against a seller or a licensed index.

What it scores, by house

BenchmarkHouseScoreEvaluatedScaffold
chess puzzlesEpoch AI0.4±0.052026-10-13xhigh
frontiermath erdosEpoch AI0.0±0.002026-09-21xhigh
frontiermath tier 4 v2Epoch AI0.2±0.062026-09-22xhigh
frontiermath tiers 1 3 v2Epoch AI0.5±0.032026-09-22xhigh
furniture assemblyEpoch AI0.2±0.052026-09-24xhigh
gpqa diamondEpoch AI0.9±0.022026-09-22xhigh
mirrorcodeEpoch AI0.5±0.112026-10-29xhigh
mystery game puzzlesEpoch AI0.3±0.052026-09-22xhigh
otis mock aime 2024 2025near its ceilingEpoch AI1.0±0.012026-09-22xhigh
simpleqa verifiedEpoch AI0.6±0.022026-09-22xhigh
ale benchcollected by Epoch AI1546—xhigh
apex agentscollected by Epoch AI0.7—unknown
arc aginear its ceilingcollected by Epoch AI0.9—medium
arc aginear its ceilingcollected by Epoch AI0.6—low
arc aginear its ceilingcollected by Epoch AI0.9—xhigh
arc aginear its ceilingcollected by Epoch AI0.9—high
arc agi 2collected by Epoch AI0.6—xhigh
arc agi 2collected by Epoch AI0.6—medium
arc agi 2collected by Epoch AI0.2—low
arc agi 2collected by Epoch AI0.6—high
blueprint bench 2collected by Epoch AI0.3—unknown
critptcollected by Epoch AI0.2—high
critptcollected by Epoch AI0.1—low
critptcollected by Epoch AI0.2—xhigh
cursorbenchcollected by Epoch AI0.4—high
cursorbenchcollected by Epoch AI0.3—low
cursorbenchcollected by Epoch AI0.4—medium
cursorbenchcollected by Epoch AI0.5—xhigh
dtbenchnear its ceilingcollected by Epoch AI1.0—xhigh
frontiercodecollected by Epoch AI0.5—grok-build
frontierswecollected by Epoch AI0.3—proximus
gdp pdfcollected by Epoch AI0.2—xhigh
geobenchcollected by Epoch AI3525—unknown
lmcacollected by Epoch AI49.4—xhigh
proofbenchcollected by Epoch AI0.3—unknown
scicodecollected by Epoch AI0.6—xhigh
scicodecollected by Epoch AI0.5—low
scicodecollected by Epoch AI0.6—high
vending bench 2collected by Epoch AI10537—unknown
webdev arenacollected by Epoch AI1638—xhigh
weirdml v3collected by Epoch AI0.1—codex_cli

These scores are not comparable between houses and are not added up. A score is dated by when it was evaluated, not when the model was released, and a score without its scaffold is not a measurement — which is why both are printed.

Every rate we hold · Why there is no single best model · Everything from xAI