BenchLeader

Codex Plus: limits, models and real cost

Codex Plus costs $20 a month. It is the ChatGPT Plus subscription, which includes Codex. It runs GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna. OpenAI describes the allowance as “Power a few focused coding sessions each week; 15–160 local GPT-6.1 Sol messages per 5 hours (estimate)”. How it refreshes: 5-hour window, plus weekly limits. Used in full on GPT-6 Sol, the allowance comes to about 1.1 billion tokens a month by Real API Pricing's high-confidence estimate: about $323 at API prices, 16× the subscription.

Price
$20
a month
Allowance in dollars
$323
estimate on GPT-6 Sol
API usage per dollar
16×
at full use
Break-even
25 sessions
a month on GPT-6 Astra

About Codex

Codex usage is bundled into the ChatGPT subscription. OpenAI publishes local messages per five-hour window for Plus (GPT-6 Astra 5–45, GPT-6.1 Sol 15–160, GPT-6.1 Terra 35–285, GPT-6 Luna 350–3,000) and says these are estimates, not fixed limits. Fast mode uses the included allowance at 2.5x the Standard rate and Ultrafast at 8x. Pro comes in three tiers; Pro plans currently have no five-hour limit, only weekly limits.

Limits and reset windows

In OpenAI’s words“Power a few focused coding sessions each week; 15–160 local GPT-6.1 Sol messages per 5 hours (estimate)”
How it refreshes5-hour window, plus weekly limits
How the size is statedNot published
Five-hour limitYes
Weekly limitYes
Shared with chatFrom December 2026
Extra resets6 full resets announced by OpenAI in the 28 days to 16 Nov 2026; discretionary, with no schedule. With every one landing on a fully used week, a month would hold up to 2.50× the usual allowance.

Models, effort and what a task costs

Artificial Analysis measured GPT-6 Astra at 5 efforts. One task of its Intelligence Index costs $1.42 through the API at low effort and $6.24 at max; through Codex Plus at full use, about $0.12 and $0.53. In T3 Code's usage data an average GPT-6 Astra turn uses about 2.2 million tokens, 97% of them read back from cache, and costs $3.24 at API list price, so a full allowance holds roughly 72 such turns a month.

ModelEffortIndexOne task, APIOne task, planTasks a month
GPT-6 Astralow62.0$1.42$0.12164
GPT-6 Astramedium64.4$2.73$0.2386
GPT-6 Astrahigh66.3$3.24$0.2872
GPT-6 Astraxhigh66.0$3.66$0.3164
GPT-6 Astramax67.3$6.24$0.5337
GPT-6.1 Sollow60.0$0.16$0.0111,850
GPT-6.1 Solmedium62.7$0.36$0.024831
GPT-6.1 Solhigh63.4$0.59$0.040505
GPT-6.1 Solxhigh63.6$0.76$0.051391
GPT-6.1 Solmax66.6$1.27$0.085235
GPT-6 Solmax62.6$2.47$0.15131
GPT-6 Lunalow45.8$0.0051$0.000826,423
GPT-6 Lunamedium49.6$0.042$0.00623,206
GPT-6 Lunahigh51.1$0.073$0.0111,854
GPT-6 Lunaxhigh52.0$0.10$0.0151,300
GPT-6 Lunamax55.1$0.15$0.022890

One task is an average task of Artificial Analysis’ Intelligence Index, at each effort it was measured at. Through the plan it is the API cost scaled by what the plan charges against its estimated allowance, so it holds only if you use the whole allowance. Effort changes how fast an allowance runs out, not its size. No allowance estimate exists for GPT-6.1 Terra on this plan.

What a real agent turn costs

A turn is one prompt run to the end, every tool call and re-read of the context included, averaged over T3 Code’s users and priced at API list price. The last column divides the allowance's API value by that price.

ModelTokens per turnFrom cacheAPI price per turnTurns a month on this plan
GPT-6 Astra2.2 million97%$3.2472
GPT-6.1 Sol3.3 million97%$0.60497
GPT-6.1 Terra3.2 million98%$0.24–
GPT-6 Luna3.0 million96%$0.0462,969

Who it suits

  • Heavy users: OpenAI announced 6 full usage resets in the last 28 days, though it calls them discretionary.

Think twice for

  • All-day agent runs: a five-hour window caps each session, under a weekly limit as well.
  • Heavy ChatGPT chat use: from December 2026 chat and Codex draw on one shared allowance.
  • Value per dollar: Codex Pro 200 buys more API usage per dollar at full use (28× against 16×).

Every Codex tier

TierPriceAllowance at full usePer dollar
Codex Go$8No figure–
Codex Plus$20About $323 on GPT-6 Sol (estimate)16×
Codex Pro 100$100No figure–
Codex Pro 200$200About $5,604 on GPT-6 Astra (estimate)28×
Codex Pro 500$500No figure–

Against the closest plans

The nearest plan in price from each of three other vendors.

PlanPriceAllowance at full usePer dollarFive-hour limitBest model
Codex Plus$20About $323 on GPT-6 Sol (estimate)16×YesGPT-6 Astra
Claude Pro$20About $1,200 on Claude Opus 5.5 (estimate)60×YesClaude Opus 5.5
Google AI Pro$19.99About $285 on Gemini 3.8 Flash (estimate)14×Not statedGemini 4 Argon
Kimi Code Plus$19About $86 on Kimi K3 (scaled estimate)4.5×Not statedKimi K3.1 Pro

Head to head

What changed

  • 1 Nov 2026 — Noted that from December 2026 Plus, Pro and Business share one token-based allowance across ChatGPT and Codex, with Instant chat exempt. OpenAI has not given the day or the size of the pool. Source
  • 1 Nov 2026 — Noted that a reset restarts the usage period it refills, so the next refresh time moves; your billing date does not. Source
  • 30 Oct 2026 — Pro 100 and Pro 200 now carry allowance estimates, carried over from their former names Pro 5x and Pro 20x; Pro 100's is doubled, as OpenAI doubled that allowance on 29 October. Source
  • 30 Oct 2026 — Codex Go corrected to $8 a month (was listed at $6). Source
  • 30 Oct 2026 — Codex Pro split into its three tiers, Pro 100, Pro 200 and Pro 500 (was a single "from $120" row); per-five-hour message estimates and speed multipliers added. Source

Codex Plus: questions people ask

How much does Codex Plus cost?
Codex Plus costs $20 a month. It is sold as ChatGPT Plus; Codex usage is bundled into the ChatGPT subscription. Price read from OpenAI's pricing page on 30 Oct 2026.
What are Codex Plus's usage limits?
OpenAI describes the allowance as “Power a few focused coding sessions each week; 15–160 local GPT-6.1 Sol messages per 5 hours (estimate)”. How it refreshes: 5-hour window, plus weekly limits. It has a five-hour limit and a weekly limit. From December 2026 ChatGPT chat and Codex share one allowance. OpenAI also announced 6 full usage resets in the 28 days to 16 Nov 2026, which it calls discretionary.
Which models does Codex Plus include?
GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna. The highest-ranked is GPT-6 Astra, at 67.3 on the BenchLeader Index. Effort changes how fast the allowance runs out, not its size: one index task on GPT-6 Astra costs 4.4 times as much at max as at low.
Is Codex Plus worth it?
Used in full on GPT-6 Sol, the allowance comes to about 1.1 billion tokens a month by Real API Pricing's high-confidence estimate: about $323 at API prices, 16× the subscription. Against paying per token, it pays for itself after about 25 agent sessions a month on GPT-6 Astra. Every figure assumes the whole allowance is used; an unused month makes every task dearer.
Codex Plus or Claude Pro?
Both cost $20 a month. Claude Pro goes further per dollar: at full use it is worth about 60× its price in API usage on Claude Opus 5.5, against 16× for Codex Plus on GPT-6 Sol.

Sources

Every plan side by side, with how the estimates are made, is on the coding plans page.

Cite as: BenchLeader, “Codex Plus: limits, models and real cost”, data as of 16 Nov 2026, https://www.benchleader.com/coding-plans/codex-plus.