whatmodel.sh / Gemini 2.5 Pro Experimental (03-25)
Gemini 2.5 Pro Experimental (03-25)
↗ share
+ contribute
Gemini 2.5 Pro Experimental (03-25) is a closed-weights language model from Google, released March 2025. Licence: proprietary.
Catalogue identity notes
Comparison model from the Epoch AI internal-evaluation CSVs (Epoch release date 2025-03-25). Experimental API model gemini-2.5-pro-exp-03-25; kept separate from gemini-2.5-pro-preview-03-25 (different API id; no source in hand states they are the same checkpoint) and from GA gemini-2-5-pro. Gemini API changelog: shut down.
$ how it compares
The LLMs ranked either side of Gemini 2.5 Pro Experimental (03-25), out of 828 with a score, then more from Google.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 116 |  DeepSeek V3.2 | 68±1 | api $0.29/M +local | ||
| 117 |  MiMo-V2-Omni | 68±2 | comparison only | ||
| 118 | Gemini 2.5 Pro Experimental (03-25) | 68±3 | comparison only | ||
| 119 |  Grok 4 (0709) | 68±1 | comparison only | ||
| 120 |  MiniMax-M2.7 | 68±1 | api $0.53/M +local | ||
| more from Google | |||||
| 9 | Gemini 4 Argon (limited preview) | 82±1 | limited preview | ||
| 23 | Gemini 3.8 Flash | 77±1 | api $1.50/M | ||
| 32 | Gemini 3.7 Flash | 76±1 | api $1.50/M | ||
$ what it's good at
agent tasks▁▁▂▆█▆▅▄▄▃▂▁ #221 of 340 * · 10% of its score · 1 test
Trails SmolLM3 3B, Jamba 1.5 Mini, Mellum2 12B-A2.5B Instruct here, though they rank lower overall.
coding▁▁▁▃▅▇█▇▅▄▃▁ #56 of 524 * · 4.5% of its score · 1 test
Beats GPT-5.4, DeepSeek V4 Flash Vision (Exp), Gemini 3.1 Pro (Preview) here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LiveCodeBench official release_v5 (Nov 1, 2023–Feb 1, 2025) | 80.9% ✎ | █···█····█·█ beats 3/3 | 4.5% |
other tests3 not in a task · 83% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond | 83.8% ✎ | ▁▁▁▁▂▄▆▄▅▇█▅ beats 141/208 | 76% |
| Open PT-LLM Leaderboard average (9 Portuguese tasks) | 88.4% ✎ | ▃▁▁▁▁▁▁▁▂█▅▃ beats 276/276 | 5.7% |
| Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate) | 98.9% ✎ | ▁▁▂▃▄▄█▆▃▂▃▁ beats 136/140 | 1.5% |
$ where to run it
No verified API or download recorded yet · add one
$ how the score was worked out · Epoch AI — internal evaluatio… 76%, Berkeley Function Calling Lea… 10%, Open Portuguese LLM Leaderboa… 6% and 2 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 Epoch AI — internal evaluations | 1 | 76% |
 Berkeley Function Calling Leaderboard result archive | 1 | 10% |
| Open Portuguese LLM Leaderboard (archived) | 1 | 6% |
| LiveCodeBench official website results (2025-08-01) | 1 | 5% |
 Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository) | 1 | 1% |
| prior · pull toward the middle | 3% | |
| weighted average | = 68.1 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json