Name: Claude Code observed coding usage snapshot
Creator: Tokscale

Reality Snapshot

Observed tokens

2.9T-2.3%

over 30 days

Estimated tracked spend

$2.7M

Invoices may differ

Active developers

678-197

deduplicated over 30 days

Observed messages

13.7M

over 30 days

Most-used model

Claude Opus 531.8%

of observed tokens

Observed Usage

Observed tokensDaily active developers

Models used

Observed 30-day client usage by model

8 observed models
Model	Tokens	Share	Change
Claude Opus 5Anthropic	935.6B	32%	0%
Claude Opus 4.8Anthropic	792.5B	27%	0%
Claude Fable 5Anthropic	606.5B	21%	0%
Claude Sonnet 5Anthropic	453.8B	15%	0%
GPT-5.6 SolOpenAI	39.8B	1.4%	0%
GLM 5.2Z Ai	24.1B	<1%	0%
Claude Sonnet 4.6Anthropic	11B	<1%	0%
MiniMax M3Minimax	9.8B	<1%	0%

Token composition

Share of observed client tokens

Cache read2.8T95%

Input125.6B4.3%

Output12.5B<1%

Reasoning00%

Client Economics

Claude Opus 5$696K26%

Claude Opus 4.8$687K26%

Claude Fable 5$984K37%

Claude Sonnet 5$136K5.1%

GPT-5.6 Sol$85.4K3.2%

GLM 5.2$13.1K<1%

Claude Sonnet 4.6$7.2K<1%

MiniMax M3$804<1%

Estimated spend per 1M tracked tokens$0.90

Cache-read share95.7%

Spend basisEstimated from tracked model usage

Estimated tracked spend uses Tokscale model and provider pricing data. Actual invoices may differ.

Client Momentum

-2.3%30-day growthObserved coding client

Compared with the prior non-overlapping 30-day window

Provider footprint

Anthropic2.8T
OpenAI39.8B
Z Ai24.1B
Minimax9.8B

Client details

Canonical clientclaude

TelemetryAggregated Tokscale submissions

Current window2026-07-12 to 2026-08-10

Comparison window2026-06-12 to 2026-07-11

SpendEstimated; invoices may differ

About this client

Claude Code is Anthropic's agentic coding tool for working with codebases from the terminal.

Links & sources

TOKSCALE INSIGHT

Claude Code recorded 2.9T coding tokens across 678 developers during the latest 30-day Tokscale window.

Evidence confidenceStable observed sample