OpenAI · GPT-4o · Frontier
GPT-4o pricing and limits
OpenAI's multimodal model for text, image and audio input. Still widely deployed in production, though newer models now beat it on price for text-only work.
| Line | Tokens | Rate / 1M | Cost |
|---|---|---|---|
| Input, uncached | 5,000 | $2.50 | $0.0125 |
| Input, cache read | 3,000 | $1.25 | $0.00375 |
| Output | 700 | $10.00 | $0.00700 |
- Input / 1M
- $2.50
- Output / 1M
- $10.00
- Cached / 1M
- $1.25
- Cache write / 1M
- n/a
- Blended 3:1
- $4.38
- Context
- 128K
What drives GPT-4o cost
GPT-4o has no hidden reasoning tokens, so output cost equals visible output. Image inputs are converted to tokens based on resolution and detail level, so an app that uploads full-resolution screenshots on every turn pays far more than the text length suggests.
Cached input is billed at half the input rate, a smaller discount than the GPT-4.1 and GPT-5 families. For prompt-heavy workloads that do not need audio or vision, migrating to GPT-4.1 often cuts input cost without code changes beyond the model name.
Cost of a typical call
Take a multimodal chat turn with 8,000 input tokens, 3,000 of them served from cache, and 700 output tokens. At GPT-4o list prices ($2.50 input, $1.25 cached input, $10.00 output per 1M) one call costs $0.0232, against $0.0270 with no cache hits. At 150,000 calls a month that is $3,488, or $4,050 uncached.
Blended at a 3:1 input to output ratio, GPT-4o costs $4.38 per 1M tokens. Routing the same call to the fallback in the policy below, GPT-4o mini, costs $0.00139, which is what a tenant pays you in COGS once they cross the hard cap and PALM SOUQ downgrades instead of blocking.
Cost comparison
GPT-4o against three alternatives
The same multimodal chat turn priced on GPT-4.1, GPT-5, Gemini 2.5 Flash. Move the sliders to match your traffic; models without a cache-read rate bill every input token in full.
Benchmark quality per task before switching. A cheaper model that needs more steps or longer outputs can cost more per finished task than its rate suggests, which is why PALM SOUQ reports cost per feature, not only per token.
GPT-4o this guide
$0.0232 per call
$2,320 / mo
baseline
GPT-4.1
$0.0170 per call
$1,704 / mo
-27% vs GPT-4o
GPT-5
$0.0136 per call
$1,358 / mo
-41% vs GPT-4o
Gemini 2.5 Flash
$0.00347 per call
$347 / mo
-85% vs GPT-4o
List prices from the PALM SOUQ LLM Price Index, computed per call as uncached input + cache reads + output. Reasoning tokens count as output on reasoning models.
Rate limits by usage tier
Per-organization TPM and RPM by usage tier, with separate limits for some multimodal endpoints. Image-heavy tenants consume TPM faster than text tenants with the same request count.
Exact limits vary by account tier and change over time, so we do not reprint them here. Check the limits page for your own account in the OpenAI documentation at platform.openai.com/docs. PALM SOUQ enforces your own per-tenant limits in front of the provider's, so a single customer cannot consume the shared allowance that every other tenant depends on.
Circuit-breaker best practices
Meter image tokens per tenant separately from text tokens using the X-PalmSouq-Feature header, so you can see which features drive vision cost and set a cap on that feature alone.
A max step cap of 15 to 20 is a sensible default for GPT-4o tool agents; it rarely needs more steps than that for a single user request.
When a breaker trips, the gateway answers in under 50 ms with HTTP 429 and X-PalmSouq-Reason: circuit_breaker_triggered, before the request reaches OpenAI. Your agent framework sees a normal rate-limit response and stops, and the audit log records which heuristic fired and for which tenant.
# palmsouq.policy.yaml
policy: gpt-4o-pro
match:
model: gpt-4o
tier: pro # from X-PalmSouq-Tier
budgets:
monthly_usd: 300 # per tenant
soft_cap_pct: 80 # webhook to your app
on_hard_cap: fallback # or: block (HTTP 429)
fallback:
- gpt-4o-mini
breakers:
tool_thrashing: { identical_calls: 3, window_steps: 8 }
token_velocity: { max_tokens_per_s: 50000, window_s: 10 }
semantic_echo: { repeats: 3 }
max_steps: 18
cache:
meter_reads: true- budgetsPer-tenant monthly dollar cap with an 80% soft-cap webhook, keyed by X-PalmSouq-Tenant.
- fallbackAt the hard cap, requests reroute to GPT-4o mini ($0.26 blended per 1M) instead of failing.
- breakersAll four loop heuristics on, evaluated in the request path with under 5 ms added latency.
Where it fits
Use where multimodal input is the feature: screenshot understanding, document photos, voice. For pure text, compare against GPT-4.1 and GPT-5 mini first.
Provider documentation: platform.openai.com/docs
GPT-4o pricing questions
How much does GPT-4o cost per million tokens?
At list price GPT-4o costs $2.50 per 1M input tokens and $10.00 per 1M output tokens, with cached input at $1.25 per 1M. Prices were last reviewed in the PALM SOUQ LLM Price Index and are for the standard tier and shortest context band.
Does GPT-4o support prompt caching?
Yes. Cached input is billed at $1.25 per 1M tokens against $2.50 for uncached input. PALM SOUQ records cached tokens per tenant from the provider's usage data, so cache savings show up in each customer's COGS.
What are the GPT-4o rate limits?
Per-organization TPM and RPM by usage tier, with separate limits for some multimodal endpoints. Image-heavy tenants consume TPM faster than text tenants with the same request count. For your account's exact numbers, see the OpenAI documentation at platform.openai.com/docs.
How do I stop runaway GPT-4o agent loops?
Meter image tokens per tenant separately from text tokens using the X-PalmSouq-Feature header, so you can see which features drive vision cost and set a cap on that feature alone. The sample policy on this page enables all four PALM SOUQ heuristics with a max step cap of 18 and falls back to GPT-4o mini at the hard cap.
Is there a batch discount for GPT-4o?
50% off input and output through the Batch API (24-hour completion window). Batch traffic suits evaluations, backfills and nightly enrichment jobs that do not need a real-time answer.
Govern GPT-4o spend per customer.
Point your OpenAI client at gateway.palmsouqltd.com, pass X-PalmSouq-Tenant, and every GPT-4o token is attributed, capped and protected by circuit breakers.
14-day free production trial · 100k requests included · no card required