← All models

Gemini 2.5 Pro

Language Deprecated Full comparison ↗

google-deepmind/gemini-2.5-pro · by Google DeepMind · decoder-only-transformer

Pricing — 1 offering(s)

About this offering: Deprecated. Gemini 2.5 Pro is no longer available to select in Google AI Studio, and API access is limited to existing users who have used it before. From 2026-12-01, Google automatically routes API requests for gemini-2.5-pro to Gemini 3.8 Flash and bills them at Gemini 3.8 Flash rates, so it's priced here through 2026-11-30. Until then it costs $1.25 input and $10 output per 1M tokens for prompts up to 200K tokens, rising to $2.50 / $15 above that (Gemini API deprecations and pricing pages, checked 2026-11-03).

Input tokens

  • $1.25 / 1M tokens (input) Current 2026-06-12 → 2026-11-30

Cached input tokens (prompts <= 200k)

  • $0.125 / 1M tokens (cached input) Current 2026-06-12 → 2026-11-30

    Storage is billed separately at $4.50/1M tokens per hour.

Output tokens

  • $10.00 / 1M tokens (output) Current 2026-06-12 → 2026-11-30

Input tokens (prompts > 200k)

  • $2.50 / 1M tokens (input) Current 2026-06-12 → 2026-11-30

Cached input tokens (prompts > 200k)

  • $0.25 / 1M tokens (cached input) Current 2026-06-12 → 2026-11-30

Output tokens (prompts > 200k)

  • $15.00 / 1M tokens (output) Current 2026-06-12 → 2026-11-30

Batch

Batch input tokens

  • $0.625 / 1M tokens (input) Current 2026-11-03 → 2026-11-30

Batch output tokens

  • $5.00 / 1M tokens (output) Current 2026-11-03 → 2026-11-30

Batch input tokens (prompts > 200k)

  • $1.25 / 1M tokens (input) Current 2026-11-03 → 2026-11-30

Batch output tokens (prompts > 200k)

  • $7.50 / 1M tokens (output) Current 2026-11-03 → 2026-11-30

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

See how Gemini 2.5 Pro fits into a cost-aware routing setup

See how →

Capability profile

How Gemini 2.5 Pro rates across core capability dimensions, with the task-level evidence behind each rating.

reasoning Strong
coding Strong

SWE-bench Verified (500 human-verified GitHub issues; patch must pass the repo's own tests; no file-location hints): resolves 63.8% — fails roughly 1 issue in 3. Vendor-reported (Google DeepMind, Gemini 2.5 Pro launch; agentless scaffold).

tool use Strong
instruction following Strong
context window Strong

1M-token context — among the largest available.

multilingual Strong
speed Moderate
cost efficiency Moderate
Benchmark Score Harness Source
swe-bench-verified ↗ 63.8% 2025-03 agentless 🏢 vendor ↗
aider-polyglot ↗ 71.4% 2025-05 aider 📊 leaderboard ↗
terminal-bench-2-1 28.5% 2026-08 terminus-2 🔬 independent ↗
bigcodebench ↗ 29.7% 2025-03 — 📊 leaderboard ↗

Operator guidance

Route here when you need a very large context window or native multimodal input alongside frontier reasoning. Drop to Gemini 2.5 Flash for cost/latency.

Use cases

Limitations

Citations