OpenAI Codex OAuth (SSE)
The lab behind ChatGPT and the GPT-5.x family.
OpenAI is the AI lab that built ChatGPT. Its GPT-5.x lineup covers the full quality/speed tradeoff — from GPT-5.5 Pro for hard research-grade work down to GPT-5.4-nano for high-volume cheap calls. We benchmark via the Codex coding agent included with ChatGPT Plus and above, so the relevant cost is the monthly ChatGPT subscription you already pay.
Strengths
- Mature ecosystem, broad tool support
- Uses ChatGPT/Codex subscription quota you already pay for
- GPT-5.x line covers the full quality/speed tradeoff
When to use it
- You already pay for ChatGPT and want to reuse the quota
- Microsoft/Azure ecosystem integration
- Agentic workloads needing the broad GPT tooling baseline
Subscription plans
| Plan | Price | Quota | Available |
|---|---|---|---|
| Plus | $20/mo | Standard ChatGPT + Codex access | yes |
| Pro | $200/mo | Higher quotas + research tier | yes |
Notes: Codex is included with ChatGPT Plus, Pro, Business and Enterprise/Edu. Free and Go tiers currently get limited Codex access (promotional).
Referral: OpenAI does not run a public referral program. Eligible Plus users may occasionally receive a small batch of trial-invite codes — they're not shareable as an open link.
Models tested on OpenAI
Speed numbers below are specific to OpenAI's routing and hardware. The same model may appear on other providers' pages with different throughput.
| Model | Best tok/s | Avg tok/s | Runs | Success | Longest output (chars) |
|---|---|---|---|---|---|
| gpt-5.4-mini | 146.8 | 126.6 | 4 | 100% | 2,311 |
| gpt-5.4-mini-fast | 141.2 | 117.9 | 4 | 100% | 2,139 |
| gpt-5.5-fast | 61.0 | 48.8 | 4 | 100% | 2,121 |
| gpt-5.2 | 53.2 | 52.1 | 4 | 100% | 2,724 |
| gpt-5.3-codex | 52.7 | 49.5 | 4 | 100% | 2,534 |
| gpt-5.4-fast | 52.4 | 50.7 | 4 | 100% | 1,924 |
| gpt-5.4 | 51.9 | 46.5 | 4 | 100% | 1,908 |
| gpt-5.5 | 50.5 | 46.7 | 4 | 100% | 2,207 |