Compare observed and estimated weekly tokens, five-hour usage and API-equivalent value for GPT-6.1 Sol, GPT-6 Sol, GPT-6 Astra and GPT-5.6 Sol across Codex plans.
Models share the plan allowance; rows are not additive. Tokens include cache. Values are usage estimates.
How these values are estimated
Rows include sample extrapolations, source reports and model-ratio estimates; not every row is a full-cycle measurement. Five hours is a short window; weekly values may be extrapolated from quota percentages or window counts. Where weekly values exist, the average month uses weekly × 52 ÷ 12 and is not a measured full month.
Tokens include cache; input/output mix changes the API equivalent. API equivalents use recorded usage and reference rates for comparison, not actual bills or guaranteed allowances. Expand each plan for its sample conditions and sources; missing model data is not filled in.
OpenAI
3 plans
Plus
Monthly fee$20
Model
Weekly100M tokens
Weekly API equivalentUSD
GPT-6 Astra
≈ 0.43
≈ $66
GPT-6.1 Sol
≈ 2.38
≈ $47
GPT-6 Sol
≈ 2.59
≈ $66
GPT-5.6 Sol
≈ 1.65
≈ $86
Estimation basis and sources
GPT-6 Astra
Model-specific Plus user report · Plus five-hour user measurement; weekly estimate assumes six windows · Estimated from reference model token mix
Sep 30, 2026 · Plus community sample · Two five-hour readings: $7.5/$7.8; adopts $7.8, assuming six windows per week · Tokens estimated with 97% cached input, 2.5% uncached input and 0.5% output
Sep 30, 2026 · Existing plan-ratio estimate · Plus baseline scaled by existing GPT-6 Sol plan ratios; not measured on this tier · Tokens estimated with 97% cached input, 2.5% uncached input and 0.5% output
Sep 30, 2026 · Existing plan-ratio estimate · Plus baseline scaled by existing GPT-6 Sol plan ratios; not measured on this tier · Tokens estimated with 97% cached input, 2.5% uncached input and 0.5% output
Compare individual model rows. Quota consumption, input/output prices and cache mix differ by model, so more tokens do not necessarily mean a higher API equivalent. Models share the plan allowance; their rows are not additive.
Are the Pro 5× and 20× weekly values simple multiples?
Each row retains its own sample or estimation basis. Plan-name multipliers are not applied uniformly across models. Some rows use usage records; others use sample ratios. Expand the sources under each plan to check the calculation.
Should I compare five hours, a week or an average month?
Use five-hour records for sustained sessions and weekly records for ongoing work. Average months use weekly × 52 ÷ 12 for comparison with monthly fees, not an allowance available all at once. Plans without short-window data are omitted from the five-hour view.
