GPT-5.3 Codex Spark
Observed throughTokscale Model Report
GPT-5.3 Codex Spark model details and observed coding usage from Tokscale.
Reality Snapshot
Observed tokens
13.1B+37.8%
over 30 days
Estimated tracked spend
$5K
Invoices may differ
Active developers
124-61
this month
30-day momentum
+37.8%
Compared with the prior 30-day window
Most-used client
Codex CLI68%
of observed tokens
Observed Usage
Tokens observedDaily active developers
How developers use it
Usage by coding client
Sharevs 30 days
Token composition
Share of observed tokens
93%cache read
Cache read12.1B93%
Input824.3M6.3%
Output68.1M<1%
Reasoning36.9M<1%
Cache-heavy workloadMost observed tokens reuse cached context across requests.
Observed Variants
default
gpt-5.3-codex-spark13B$5K$0.3899.72%router via unknown
router_client:unknown/gpt-5.3-codex-spark36.2M$0$0.000.28%Token Economics
Estimated spend profile
Codex CLI$3.1K63.14%
Pi$75415.17%
OpenCode$91618.45%
Hermes Agent$951.9%
Claude Code$591.18%
Synthetic$0.000%
OpenClaw$7.750.16%
Goose$0.020%
Observed cost signals
Estimated spend per 1M tracked tokens$0.38
Cache-read share92.9%
Estimated cache savingsNot available
Typical workloadcontext-heavy
Estimated tracked spend calculated from Tokscale model and provider pricing data. Actual invoices may differ.
Model Momentum
+37.8%30-day growthRank: Observed model
Compared with the prior non-overlapping window
Fastest-growing client: Codex CLI
Official Model Details
ProviderOpenAI
Tokscale Insight
GPT-5.3 Codex Spark recorded 13.1B coding tokens across 124 developers during the latest 30-day Tokscale window.
Observed confidenceHigh