Google DeepMind · The Reasoning-Model Era
Gemini 2.5 Pro
Google's reasoning flagship. Two-million-token context, native multimodal, the only frontier model that reads PDFs without an extraction pre-pass.
By C.W. Jameson · Published 19 May 2026 · Last reviewed 19 May 2026
Gemini 2.5 Pro has the largest production context window of any frontier model and the most aggressive multimodal training. It can ingest a one-hour video, a two-hundred-page PDF, and forty images in a single call, and answer questions about all of them. The pricing is competitive at small scale and punishing at long-context scale. Operators who measured ended up running it primarily for the unique workloads no other model can attempt.
Field signature
Will answer questions about specific frames of long videos. Other models can't.
Specifications
| Released | 2025 |
|---|---|
| Context window | 2,000,000 tokens |
| Pricing | $1.25 / $10 input/output up to 200K, doubles past that |
| Modalities | text · image · audio · video · PDF |
| License | Commercial API only |
| Era | The Reasoning-Model Era |
Strengths
- Context length
- Native video understanding
- Aggressive multimodal
Weaknesses
- Long-context pricing
- Inconsistent tool-use compared to Claude
Authentication markers
The fingerprints by which Gemini 2.5 Pro can be identified from its output alone.
| Tell | Meaning |
|---|---|
| Can answer 'what is at timestamp 23:14 in the video?' correctly. | Gemini 2.5 lineage. |
Notable works
- AlphaCode 2
- AlphaFold 3 (related lineage)
Market position
$1.25 - $20 per million tokens
Partner offer
Gemini's long-context window has no peer at the time of this entry.
Try Gemini →Affiliate link — see disclosure.
Primary sources
From the Almanac shop
The Operator's Compendium
Every agent harness, every routing pattern, every cost trick. 90-page PDF.
$29 — Coming soon