OpenAI · GPT-4.1 · Frontier
GPT-4.1 pricing and limits
OpenAI's non-reasoning flagship with a 1M token context window. No hidden thinking tokens, so cost is predictable from prompt and completion length, and long documents do not carry a context surcharge.
| Line | Tokens | Rate / 1M | Cost |
|---|---|---|---|
| Input, uncached | 10,000 | $2.00 | $0.0200 |
| Input, cache read | 50,000 | $0.50 | $0.0250 |
| Output | 800 | $8.00 | $0.00640 |
- Input / 1M
- $2.00
- Output / 1M
- $8.00
- Cached / 1M
- $0.50
- Cache write / 1M
- n/a
- Blended 3:1
- $3.50
- Context
- 1M
What drives GPT-4.1 cost
With a 1M context window, the risk is not the per-token price but the prompt size your application lets through. A retrieval step that stuffs an entire repository or contract archive into context can bill hundreds of thousands of input tokens per call, and every follow-up turn repeats them.
Cached input is billed at a quarter of the input rate. For long-document chat, where each turn resends the same document, cache hits are the difference between a viable feature and a margin problem. Keep the document at the start of the prompt and the conversation after it.
Cost of a typical call
Take a long-document Q&A turn with 60,000 input tokens, 50,000 of them served from cache, and 800 output tokens. At GPT-4.1 list prices ($2.00 input, $0.50 cached input, $8.00 output per 1M) one call costs $0.0514, against $0.1264 with no cache hits. At 90,000 calls a month that is $4,626, or $11,376 uncached.
Blended at a 3:1 input to output ratio, GPT-4.1 costs $3.50 per 1M tokens. Routing the same call to the fallback in the policy below, GPT-4.1 mini, costs $0.0103, 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-4.1 against three alternatives
The same long-document Q&A turn priced on GPT-5, Gemini 2.5 Pro, Claude Sonnet 4.5. 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-4.1 this guide
$0.0517 per call
$5,170 / mo
baseline
GPT-5
$0.0270 per call
$2,698 / mo
-48% vs GPT-4.1
Gemini 2.5 Pro
$0.0362 per call
$3,619 / mo
-30% vs GPT-4.1
Claude Sonnet 4.5
$0.0575 per call
$5,754 / mo
+11% vs GPT-4.1
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 tokens per minute and requests per minute, scaled by usage tier. Long-context requests consume a large share of a TPM budget in a single call, so a handful of tenants uploading big files can throttle everyone else unless you rate limit per tenant.
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
Add a per-request input token ceiling per tier. PALM SOUQ can reject or downgrade requests whose prompt exceeds a size you choose before they reach OpenAI, which protects you from runaway context growth in multi-turn sessions.
Watch for context-window overflow loops, where an agent summarizes its own history, appends the summary, and summarizes again. The recursion depth cap and semantic echo check both catch this pattern.
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-4-1-business
match:
model: gpt-4.1
tier: business # from X-PalmSouq-Tier
budgets:
monthly_usd: 600 # per tenant
soft_cap_pct: 80 # webhook to your app
on_hard_cap: fallback # or: block (HTTP 429)
fallback:
- gpt-4.1-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: 15
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-4.1 mini ($0.70 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
Strong for long-document question answering, code review over large diffs and instruction-following tasks that do not need deliberate reasoning. Predictable cost per call makes it easy to price into a plan.
Provider documentation: platform.openai.com/docs
GPT-4.1 pricing questions
How much does GPT-4.1 cost per million tokens?
At list price GPT-4.1 costs $2.00 per 1M input tokens and $8.00 per 1M output tokens, with cached input at $0.50 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-4.1 support prompt caching?
Yes. Cached input is billed at $0.50 per 1M tokens against $2.00 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-4.1 rate limits?
Per-organization tokens per minute and requests per minute, scaled by usage tier. Long-context requests consume a large share of a TPM budget in a single call, so a handful of tenants uploading big files can throttle everyone else unless you rate limit per tenant. For your account's exact numbers, see the OpenAI documentation at platform.openai.com/docs.
How do I stop runaway GPT-4.1 agent loops?
Add a per-request input token ceiling per tier. PALM SOUQ can reject or downgrade requests whose prompt exceeds a size you choose before they reach OpenAI, which protects you from runaway context growth in multi-turn sessions. The sample policy on this page enables all four PALM SOUQ heuristics with a max step cap of 15 and falls back to GPT-4.1 mini at the hard cap.
Is there a batch discount for GPT-4.1?
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-4.1 spend per customer.
Point your OpenAI client at gateway.palmsouqltd.com, pass X-PalmSouq-Tenant, and every GPT-4.1 token is attributed, capped and protected by circuit breakers.
14-day free production trial · 100k requests included · no card required