Claude Ultravs
Codex Pro 200
Both cost $200 a month. Claude Ultra goes further per dollar: at full use it is worth about 62× its price in API usage on Claude Opus 5.5, against 28× for Codex Pro 200 on GPT-6 Astra.
Side by side
| Claude Ultra | Codex Pro 200 | |
|---|---|---|
| Price | $200 a month | $200 a month |
| Billed annually | – | – |
| Models | Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5 | GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna |
| Best model | Claude Opus 5.5(68.4) | GPT-6 Astra(67.3) |
| Allowance, in the vendor's words | “10x Pro usage every week; $200/month in API credits for the Claude Platform” | “More usage than Pro 100; no Ultrafast” |
| How it refreshes | Weekly limit only once the account is verified (no five-hour limit), plus a separate weekly Fable limit | Weekly limits; no five-hour limit |
| Five-hour limit | None once verified | No |
| Weekly limit | Yes | Yes |
| Shared with chat | Yes | From December 2026 |
| Allowance at full use | About $12,426 on Claude Opus 5.5 (estimate) | About $5,604 on GPT-6 Astra (estimate) |
| API usage per dollar | 62× | 28× |
| One index task through the plan | $0.16 on Claude Opus 5.5 (max), $10.04 via the API | $0.22 on GPT-6 Astra (max), $6.24 via the API |
| Average real agent turn | $1.02 on Claude Opus 5.5, about 12,224 a month | $3.24 on GPT-6 Astra, about 1,731 a month |
| Beats the API after | 625 agent sessions a month on Claude Opus 5.5 | 250 agent sessions a month on GPT-6 Astra |
| API credits included | $200 a month | – |
| Extra resets, last 28 days | – | 6 (OpenAI, discretionary) |
| Prices read | claude.com, 30 Oct 2026 | learn.chatgpt.com, 30 Oct 2026 |
Choose Claude Ultra for
- More API usage per dollar at full use: 62× its price against 28×.
- The stronger best model: Claude Opus 5.5 (BenchLeader Index 68.4) against GPT-6 Astra (67.3).
- $200 a month of Anthropic API credits on top, outside the usage pool.
Choose Codex Pro 200 for
- More models to choose from: 4 against 3.
- Extra resets: OpenAI announced 6 full usage resets in the last 28 days.
At the same effort
The same model costs several times more per task at max effort than at low, so plans only compare fairly at the same setting. Each cell is one Artificial Analysis index task on the plan’s best-supported model at that effort, through the plan at full use.
| Effort | Claude Ultra | Codex Pro 200 |
|---|---|---|
| low | $0.016 on Claude Opus 5.5, 12,738 tasks a month | $0.051 on GPT-6 Astra, 3,936 tasks a month |
| medium | $0.039 on Claude Opus 5.5, 5,070 tasks a month | $0.098 on GPT-6 Astra, 2,051 tasks a month |
| high | $0.056 on Claude Opus 5.5, 3,586 tasks a month | $0.12 on GPT-6 Astra, 1,730 tasks a month |
| xhigh | $0.097 on Claude Opus 5.5, 2,055 tasks a month | $0.13 on GPT-6 Astra, 1,531 tasks a month |
| max | $0.16 on Claude Opus 5.5, 1,238 tasks a month | $0.22 on GPT-6 Astra, 897 tasks a month |
More on each plan
Claude Ultra vs Codex Pro 200: questions people ask
- Which is better for coding, Claude Ultra or Codex Pro 200?
- Both cost $200 a month. Claude Ultra goes further per dollar: at full use it is worth about 62× its price in API usage on Claude Opus 5.5, against 28× for Codex Pro 200 on GPT-6 Astra.
- Is Claude Ultra cheaper than Codex Pro 200?
- No, both cost $200 a month.
- Which gives more usage, Claude Ultra or Codex Pro 200?
- In API dollars at full use: Claude Ultra about $12,426 a month (estimated on Claude Opus 5.5); Codex Pro 200 about $5,604 a month (estimated on GPT-6 Astra). Usage is not finished work: heavier models and higher effort use an allowance up sooner.
- Which models do Claude Ultra and Codex Pro 200 include?
- Claude Ultra: Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5. Codex Pro 200: GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna.
Sources
Claude Ultra
- Price and the allowance wording: claude.com, read 30 Oct 2026.
- Limits: Claude usage limits, read 30 Oct 2026.
- Limits: Max and Ultra announcement, read 30 Oct 2026.
- Allowance estimate: Real API Pricing (MIT licence), data to 27 Sep 2026, priced at BenchLeader’s API prices of 16 Nov 2026.
- Cost per index task at each effort: Artificial Analysis, as of 16 Nov 2026.
- Real agent turns: T3 Code’s usage telemetry as published by Slopalytics, priced at API list price.
Codex Pro 200
- Price and the allowance wording: learn.chatgpt.com, read 30 Oct 2026.
- Limits: About ChatGPT Pro tiers, read 30 Oct 2026.
- Limits: How resets work in Codex, read 30 Oct 2026.
- Shared allowance from December: OpenAI help centre.
- Allowance estimate: Real API Pricing (MIT licence), data to 21 Sep 2026, priced at BenchLeader’s API prices of 16 Nov 2026.
- Cost per index task at each effort: Artificial Analysis, as of 16 Nov 2026.
- Real agent turns: T3 Code’s usage telemetry as published by Slopalytics, priced at API list price.
Estimates convert token counts to dollars at each model’s API prices with Real API Pricing’s agentic mix, so a plan estimated on a cheaper model looks smaller in dollars than the same tokens would on a dearer one.
Cite as: BenchLeader, “Claude Ultra vs Codex Pro 200”, data as of 16 Nov 2026, https://www.benchleader.com/coding-plans/claude-ultra-vs-codex-pro-200.