GPT-6.1 Sol pricing: real cost per million tokens

Plan

ChatGPT Plus · OpenAI

Data date

2026-09-30 · sampled

Real price / MTok

$0.0229885

Monthly tokens

870,000,000 (0.87B)

Quota confidence

High

Calculation

$20 / month ÷ 870,000,000 tokens × 1,000,000 ≈ $0.0229885 / MTok

Quota basis

Standard workload: 97% cache reads / 2.5% input / 0.5% output

Benchmark mapping

Artificial Analysis Intelligence Index v4.3: Medium (5 configurations)
Terminal-Bench 4.0 (AA): Low (5 configurations)
Code Arena | WebDev Overall: Low (1 configuration)
Artificial Analysis Coding Agent Index v1.5: Low (5 configurations)
OpenDesign Arena · Quality ranking: Low (1 configuration)Mapping confidence: how directly each score refers to the model served here.

Adoption evidence · source and rationale (translated from the Chinese record)

Maintainer's local Codex measurement: weekly-window remaining 81%→50% (31pp); the whole segment is gpt-6.1-sol total 55,532,850 tokens (input 2,742,389 / cache_read 52,536,064 / cache_write 0 / output 254,397 incl. reasoning 72,105, hit 95.04%); chatgpt-gpt61sol-plus-round1/round2/round3-2026-09-30.json; workload-conversion-round1-2026-09-30.json

775M→870M (2026-09-30 decision: uniform workload-normalization for measured samples with token breakdown, recomputed as a whole after folding in the third segment 60%→50%): segment worth $13.2824 (cache 52.54M×$0.10 + input 2.74M×$2 + output 0.25M×$10) ÷31%×4 weeks = $171.39 ÷ standard-tier blended price $0.197/MTok; GPT-6.1 Sol official prices $0.10/$2/$10 (learn.chatgpt.com credits 2.5/50/250 ÷25, cached = 5% of input — replacing the 896M that first assumed the 6-Sol $0.2 price; if the 6-Sol price still applied the result would be 814M, kept for comparison); the raw-basis 7.17 is kept for comparison. Doubts (2026-10-01 decision: record first, adopt per the whole segment 81%→50%): the third segment 60%→50% yields $0.330 per 1%, about 30% below the first two segments' $0.476/$0.473, which the 10pp integer-reading rounding (about ±10%) cannot explain — cause unknown. The cache-read price was confirmed on 2026-10-01: GPT-6.1 Sol $0.10, GPT-6 Sol $0.20. Original adoption decision — new 717M: whole segment 55,532,850÷31%×4 weeks (= 81%→65% segment 30,109,568÷16pp + 65%→60% segment 10,562,908÷5pp + 60%→50% segment 14,860,374÷10pp — the three segment boundaries join and the component sums match item by item); panel integer-% readings Δpp 30~32 map to 843M~899M normalized; the screenshot title's 82%/32pp starting reading was corrected to 81%; back-deriving the weekly pool from the 6.1 official credits tier gives 1,071cr ($42.85), while linux.do observes Plus at about 3,426cr/week and the Pro 20x same-pool Sol-worth÷20 is closer to $119.6/week at a 5.6 equivalent rate — a possible difference in in-subscription metering rate or pool basis is recorded as uncertainty; segment densities (6.1 prices): seg1 $0.476/pp, seg2 $0.473/pp, seg3 $0.330/pp; Pro 5x/20x not derived (follows the Sol ruling); no benchmark score yet — a supplement will be added once one exists

workload-conversion-round1-2026-09-30.jsonBrowse the public evidence archive