GPT-4o
OpenAI's most popular chat model. Every listing here is a gpt-4o variant sold through the OpenRouter hub.
Cheapest
Lowest floor price across matching listings, compared only within the same pricing unit — different-unit or unit-less rows are excluded, never converted. Per-1M-token rows take the min of input & output; platform rows also show the official maker price and the % difference.
- 2 OpenRouter openai/gpt-4o-mini official $0.150 · 0.0% vs official$0.150
- 3 OpenRouter openai/gpt-4o-mini-2024-07-18$0.150
- 4 OpenAI gpt-4o-mini$0.150
- 5 OpenRouter openai/gpt-4o:batch official $2.50 · -50.0% vs official$1.25
- + 6 more listings
Free tier
Listings that cost nothing — a $0 price or an official free-tier allowance (free_tier field).
Data incoming
No public snapshot for this ranking yet. Check back after the next daily feed update.
Fastest
Lowest p50 latency in ms. Approximated at the provider endpoint, not per model.
Most reliable
Highest share of 200 responses (2xx) from the latest probe snapshot.
Overall
Weighted score out of 100 — price 40%, reliability 30%, latency 30% (missing dims skipped & renormalized).
- 2 OpenAI price 0 · reliability 0 · latency 00/100
Platform details
per provider| provider | model | Input /1M | Output /1M | Latency ? | Official $/1M | Availability | Data time |
|---|---|---|---|---|---|---|---|
| OpenAI Official | gpt-4o-2024-05-13 | $5.00 | $15.00 | — | $5.00 | — | 2026-10-14 00:43:42 |
| OpenRouter | openai/gpt-4o-2024-05-13 | $5.00 | $15.00 | 128.1 ms | $5.00 | 100% available | 2026-10-14 00:43:21 |
2 rows · snapshot 2026-10-14 07:19:02 · from data/public/prices.json + measurements.json
How to read this page
All rankings are computed from public daily snapshots — verify any result against the source feed below.
- Cheapest ranks the floor price per 1M tokens — the lower of input and output — across every matching listing. Where a platform (e.g. OpenRouter) resells a model its maker also sells, the row shows the official maker price and the % difference — platform price vs official price (negative means the platform is cheaper).
- Fastest and Most reliable are approximated at the provider endpoint (p50 latency and 2xx share of probes), not per model.
- Overall score weights: price 40% · reliability 30% · latency 30%. Missing dims are skipped and renormalized; fewer than two dims means no overall score.
Feeds: prices.json · measurements.json · open_data.json · snapshot Oct 14, 2026, 07:19 AM