Google · Gemini 2.5 · Frontier
Gemini 2.5 Pro pricing and limits
Google's flagship thinking model with a 1M token context window. Pricing is tiered by prompt length, and thinking tokens are billed as output, so long-context agents need explicit guardrails.
| Line | Tokens | Rate / 1M | Cost |
|---|---|---|---|
| Input, uncached | 20,000 | $1.25 | $0.0250 |
| Input, cache read | 60,000 | $0.31 | $0.0186 |
| Output | 3,000 | $10.00 | $0.0300 |
- Input / 1M
- $1.25
- Output / 1M
- $10.00
- Cached / 1M
- $0.31
- Cache write / 1M
- n/a
- Blended 3:1
- $3.44
- Context
- 1M
What drives Gemini 2.5 Pro cost
Gemini 2.5 Pro prices prompts in two bands: requests up to 200K tokens use the base rate in our index, and longer prompts are billed at a higher rate for both input and output. A chat that keeps appending history crosses that line without any code change.
Thinking is always on for 2.5 Pro and thinking tokens are billed at the output rate. Implicit caching discounts repeated prefixes automatically, and explicit context caching adds an hourly storage charge for the cached tokens, which only pays off when the cache is read often enough before it expires.
Cost of a typical call
Take a long-context agent step with 80,000 input tokens, 60,000 of them served from cache, and 3,000 output tokens. At Gemini 2.5 Pro list prices ($1.25 input, $0.31 cached input, $10.00 output per 1M) one call costs $0.0736, against $0.1300 with no cache hits. At 50,000 calls a month that is $3,680, or $6,500 uncached.
Blended at a 3:1 input to output ratio, Gemini 2.5 Pro costs $3.44 per 1M tokens. Routing the same call to the fallback in the policy below, Gemini 2.5 Flash, costs $0.0180, which is what a tenant pays you in COGS once they cross the hard cap and PALM SOUQ downgrades instead of blocking.
Cost comparison
Gemini 2.5 Pro against three alternatives
The same long-context agent step priced on GPT-5, Claude Sonnet 4.5, GPT-4.1. 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.
Gemini 2.5 Pro this guide
$0.0736 per call
$3,680 / mo
baseline
GPT-5
$0.0625 per call
$3,125 / mo
-15% vs Gemini 2.5 Pro
Claude Sonnet 4.5
$0.1230 per call
$6,150 / mo
+67% vs Gemini 2.5 Pro
GPT-4.1
$0.0940 per call
$4,700 / mo
+28% vs Gemini 2.5 Pro
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
Gemini API limits are set per project as requests per minute, tokens per minute and requests per day, and rise as the project moves from the free tier through paid tiers based on billing history. Vertex AI uses separate quotas per region.
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 Google documentation at ai.google.dev/gemini-api/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
Set a per-request input ceiling below the 200K band for tiers that should never pay long-context rates, and let PALM SOUQ downgrade or reject requests above it.
Use context-window overflow detection: agents that re-summarize their own history to stay under a limit tend to loop, and the semantic echo heuristic catches repeated summaries.
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 Google. 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: gemini-2-5-pro-business
match:
model: gemini-2.5-pro
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:
- gemini-2.5-flash
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: 20
max_reasoning_tokens: 16000 # 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 Gemini 2.5 Flash ($0.85 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
Long-document analysis, video and multimodal reasoning, and agents that benefit from a large working context. Price against GPT-4.1 for long-context work that does not need thinking.
Provider documentation: ai.google.dev/gemini-api/docs
Gemini 2.5 Pro pricing questions
How much does Gemini 2.5 Pro cost per million tokens?
At list price Gemini 2.5 Pro costs $1.25 per 1M input tokens and $10.00 per 1M output tokens, with cached input at $0.31 per 1M. Prices were last reviewed in the PALM SOUQ LLM Price Index and are for the standard tier and shortest context band.
Does Gemini 2.5 Pro support prompt caching?
Yes. Cached input is billed at $0.31 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 Gemini 2.5 Pro rate limits?
Gemini API limits are set per project as requests per minute, tokens per minute and requests per day, and rise as the project moves from the free tier through paid tiers based on billing history. Vertex AI uses separate quotas per region. For your account's exact numbers, see the Google documentation at ai.google.dev/gemini-api/docs.
How do I stop runaway Gemini 2.5 Pro agent loops?
Set a per-request input ceiling below the 200K band for tiers that should never pay long-context rates, and let PALM SOUQ downgrade or reject requests above it. The sample policy on this page enables all four PALM SOUQ heuristics with a max step cap of 20 and falls back to Gemini 2.5 Flash at the hard cap.
Is there a batch discount for Gemini 2.5 Pro?
50% off through Gemini Batch Mode for asynchronous jobs. Batch traffic suits evaluations, backfills and nightly enrichment jobs that do not need a real-time answer.
Govern Gemini 2.5 Pro spend per customer.
Point your Google client at gateway.palmsouqltd.com, pass X-PalmSouq-Tenant, and every Gemini 2.5 Pro token is attributed, capped and protected by circuit breakers.
14-day free production trial · 100k requests included · no card required