← Back to directory
Google DeepMind
Gemini 1.5 Pro
Released Feb 15, 2024 · knowledge cutoff 2023-11
- Status
- Deprecated
- Location
- United States / United Kingdom
- Modality
- Multimodal
- Context window
- 2M
- Max output
- 8K
- Speed
- —
- Price ($/MTok in / out)
- $1.25 / $5
- Cost per task
- —
- Open weights
- No
Benchmarks4 of 8 reported
MMLU-Pro
69%
GPQA Diamond
46.2%
Humanity's Last Exam
4.6%
LMArena Elo
1302
SWE-bench Verified
—
Terminal-Bench 2.1
—
AIME (latest)
—
ARC-AGI-2
—
Strengths & weaknesses
Strengths
- Genuinely uses its context — finds a single line in a book-length prompt
- Watches video and listens to audio natively instead of working from transcripts
- Quality holds at input lengths where rivals degraded badly
Weaknesses
- Mid-pack analysis — the long context couldn't make up for shallow reasoning
- Safety filters fired often on plainly benign prompts
- Very slow to first token on large prompts
For developers
API model strings
Not researched
Licence
Proprietary — weights not released
Retirement
Retirement date not researched
Lineage
No recorded predecessor