RealRun / Guides

GPT vs Gemini vs Grok for receipts

Short answer: GPT won. GPT-6.1 Sol scored 95.8 with no severe errors. Gemini 4 Argon averaged 93.6 but swung from 84.1 to 98.8 between passes. Grok 4.7 scored 83.1 and was the most expensive model tested; Grok 4.3 (69.2) invented totals on unreadable receipts.

Model	Score	Held-back 20	Severe errors	Cost
Gemini 4 Argon	93.6	96.2	1 of 3 passes	medium
Gemini 3.5 Flash-Lite	85.4	87.6	both passes	low
Grok 4.7	83.1	96.1	both passes	high
Grok 4.3	69.2	89.1	every pass	medium

By family

GPT: Both GPT models (Sol and Luna) had zero severe errors. Luna is the budget pick.
Gemini: Argon is strong but inconsistent. Flash-Lite caught only a third of receipts that didn't add up.
Grok: Grok 4.7 scored 96.1 on the held-back set but 81.6 on the main set. Grok 4.3 is not recommended.

Claude models were tested too: Claude Haiku 92.8, Claude Opus 89.7. Full results.

Want the exact prompt and setup? The Receipts AI Playbook has the word-for-word tested prompt, model picks by budget, and every miss explained.

Get the playbook, $29 or get the free weekly AI test report

More from RealRun

Scores reflect our tests on the dates shown, not a guarantee. Claude models ran on a Claude subscription, so they have no metered API cost. cost_tier is relative within a task (low/medium/high) from our run spend; 'subscription' = not metered. Tested October 2026 by RealRun (Base Ledger IP LLC).