72Trust
Highly Accurate
🔍 Web Verified
Simon WillisononMastodon1d ago
I upgraded my Claude token counter tool to compare different models and Opus 4.7 does appear to use 1.46x times the tokens for text and up to 3x the tokens for images - it's priced the same as Opus 4.6 on a per-token basis so this is actually a pretty big price bump https://simonwillison.net/2026/Apr/20/claude-token-counts/
Trust Metrics
72
80
68
55
Claim Accuracy72%
Source Quality80%
Framing & Tone68%
Context55%
Analysis Summary
Claude Opus 4.7 uses significantly more tokens than Opus 4.6 (1.46x for text, possibly 3x for images) while maintaining identical per-token pricing, effectively increasing costs for the same computational work. Independent sources confirm the model pricing and note the token increase, though the exact image multiplier remains unconfirmed. Users switching to Opus 4.7 should expect higher API bills despite no official price change.
Claims Analysis (4)
“Opus 4.7 does appear to use 1.46x times the tokens for text”
VentureBeat reports 1.0–1.35x increase. Poster's 1.46x figure is higher but within plausible range of variation across different content types.
“up to 3x the tokens for images”
No independent source confirms the 3x figure for images specifically. Sources mention image efficiency improvements but not quantified at 3x.
“it's priced the same as Opus 4.6 on a per-token basis”
Multiple sources (VentureBeat, Mashable, llm-stats) confirm identical $5/$25 pricing between models.
“this is actually a pretty big price bump”
Interpretation based on verified pricing/token facts. Characterization as 'big' is reasonable analysis but not independently verifiable as objective fact.
Was this analysis helpful?
Try ClearFeed free →