OpenAI · GPT-5 · Frontier
GPT-5 pricing and limits
OpenAI's flagship reasoning model for agents and coding. Input is cheap relative to its capability, but every hidden reasoning token is billed at the output rate, so the effort setting decides your bill more than prompt length does.
| Line | Tokens | Rate / 1M | Cost |
|---|---|---|---|
| Input, uncached | 6,000 | $1.25 | $0.00750 |
| Input, cache read | 18,000 | $0.13 | $0.00225 |
| Output | 3,000 | $10.00 | $0.0300 |
- Input / 1M
- $1.25
- Output / 1M
- $10.00
- Cached / 1M
- $0.13
- Cache write / 1M
- n/a
- Blended 3:1
- $3.44
- Context
- 400K
What drives GPT-5 cost
GPT-5 thinks before it answers, and those reasoning tokens never appear in the response text yet are billed as output tokens. A request that returns 400 visible tokens can bill several thousand output tokens at high reasoning effort. The reasoning_effort parameter (minimal, low, medium, high) is the single biggest cost lever, followed by the verbosity setting.
Cached input is billed at a tenth of the input rate, and OpenAI applies caching automatically once a prompt prefix reaches 1,024 tokens. Agents that resend a long, stable system prompt and tool schema on every step benefit the most, as long as the stable part stays at the front of the prompt and nothing volatile (timestamps, request IDs) is injected above it.
Cost of a typical call
Take a coding agent step with 24,000 input tokens, 18,000 of them served from cache, and 3,000 output tokens. At GPT-5 list prices ($1.25 input, $0.13 cached input, $10.00 output per 1M) one call costs $0.0398, against $0.0600 with no cache hits. At 120,000 calls a month that is $4,770, or $7,200 uncached.
Blended at a 3:1 input to output ratio, GPT-5 costs $3.44 per 1M tokens. Routing the same call to the fallback in the policy below, GPT-5 mini, costs $0.00795, 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-5 against three alternatives
The same coding agent step priced on GPT-5 mini, Claude Sonnet 4.5, Gemini 2.5 Pro. 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-5 this guide
$0.0398 per call
$3,975 / mo
baseline
GPT-5 mini
$0.00795 per call
$795 / mo
-80% vs GPT-5
Claude Sonnet 4.5
$0.0684 per call
$6,840 / mo
+72% vs GPT-5
Gemini 2.5 Pro
$0.0431 per call
$4,308 / mo
+8% vs GPT-5
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
OpenAI sets tokens per minute and requests per minute per organization and per model, and both rise automatically as your account moves through usage tiers based on cumulative spend and account age. New organizations start on the lowest tier, so a launch day spike can hit 429s long before any budget is touched.
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
Cap reasoning per request: PALM SOUQ reads the reasoning token count from the usage block on every response and can trip when a single tenant's reasoning tokens per step exceed a ceiling you set, which catches agents that loop on deliberation rather than on tool calls.
Pair a max step cap with tool thrashing detection. GPT-5 agents rarely repeat themselves word for word; they more often retry the same tool with slightly altered arguments, which the oscillating argument check is built to spot.
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-5-pro
match:
model: gpt-5
tier: pro # from X-PalmSouq-Tier
budgets:
monthly_usd: 400 # per tenant
soft_cap_pct: 80 # webhook to your app
on_hard_cap: fallback # or: block (HTTP 429)
fallback:
- gpt-5-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: 30
max_reasoning_tokens: 12000 # per request
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-5 mini ($0.69 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
Best for complex agentic workflows where quality per step matters more than price per token. Route routine classification and extraction to GPT-5 mini or nano and keep GPT-5 for planning and hard reasoning steps.
Provider documentation: platform.openai.com/docs
GPT-5 pricing questions
How much does GPT-5 cost per million tokens?
At list price GPT-5 costs $1.25 per 1M input tokens and $10.00 per 1M output tokens, with cached input at $0.13 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-5 support prompt caching?
Yes. Cached input is billed at $0.13 per 1M tokens against $1.25 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-5 rate limits?
OpenAI sets tokens per minute and requests per minute per organization and per model, and both rise automatically as your account moves through usage tiers based on cumulative spend and account age. New organizations start on the lowest tier, so a launch day spike can hit 429s long before any budget is touched. For your account's exact numbers, see the OpenAI documentation at platform.openai.com/docs.
How do I stop runaway GPT-5 agent loops?
Cap reasoning per request: PALM SOUQ reads the reasoning token count from the usage block on every response and can trip when a single tenant's reasoning tokens per step exceed a ceiling you set, which catches agents that loop on deliberation rather than on tool calls. The sample policy on this page enables all four PALM SOUQ heuristics with a max step cap of 30 and falls back to GPT-5 mini at the hard cap.
Is there a batch discount for GPT-5?
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-5 spend per customer.
Point your OpenAI client at gateway.palmsouqltd.com, pass X-PalmSouq-Tenant, and every GPT-5 token is attributed, capped and protected by circuit breakers.
14-day free production trial · 100k requests included · no card required